Dialogue qualification funnel

Every cloud profile tested through Rundale's production dialogue path. Structural reliability, guard use, and request reliability are hard gates; player-facing latency ranks the profiles that survive.

38 runs1 invalid profiles15 rejected1 stopped10 quality qualified3 quality rejected3 awaiting judgment5 awaiting adjudication

Screening policy

Reliability is gated. Speed is ranked among the survivors.

immutable cloud-dialogue qualification receipts

Preflight12 calls

Valid responses100.0%

Guard interventions≤ 10.0%

Error rate≤ 0.5%

Performance sample15+ total

Cold / warm sample5 / 10+

Latencyranking only

Speed index75% TTFT + 25% completion

Judge panel3 families · median

Independent votes2+ · self-family excluded

Quality overall≥ 3.5

Critical axes≥ 3.0

Hard failures≤ 0

Blinded dialogue quality ranking

Blinded multi-family consensus · self-family scores excluded · 18 retained production outputs per judge

Rank	Model	Verdict	Overall	Minimum critical axis	Hard failures	Unusable outputs	Speed rank	Judge cost
#1	`openai/gpt-5.6-luna`	qualified	4.56	4.44	0	0	#11	$0.909
#2	`gemini-3.7-flash`	qualified	4.54	4.47	0	0	#16	$1.407
#3	`google/gemini-3.6-flash`	qualified	4.45	4.08	0	0	#9	$1.407
#4	`google/gemini-3.7-flash`	qualified	4.44	4.06	0	0	#15	$1.463
#5	`openai/gpt-5.6-luna`	rejected	4.43	4.00	1	1	#10	$1.798
#6	`google/gemini-3.7-flash`	qualified	4.42	4.11	0	0	#2	$1.465
#7	`google/gemini-3.5-flash-lite`	qualified	4.41	4.17	0	0	#6	$1.452
#8	`gemini-3.7-flash`	qualified	4.34	4.00	0	0	#4	$1.395
#9	`moonshotai/kimi-k2-0905`	qualified	4.34	3.89	0	0	#5	$1.791
#10	`gemini-3.7-flash`	qualified	4.33	4.08	0	0	#14	$1.396
#11	`google/gemini-3.7-flash`	qualified	4.33	4.03	0	0	#13	$1.416
#12	`openai/gpt-4.1-mini`	rejected	3.77	2.89	6	0	#3	$1.780
#13	`qwen/qwen2.5-vl-72b-instruct`	rejected	3.71	3.00	9	0	#17	$1.715

Qualified cloud speed ranking

Lower speed index is better. Quality judgment still determines whether dialogue is usable.

Rank	Model	Reasoning	Max tokens	Warm TTFT p95	Warm completion p95	Speed index	Error rate
#1	`moonshotai/kimi-k2-0905:nitro`	provider default	768	1336 ms	1344 ms	1338 ms	0.0%
#2	`google/gemini-3.7-flash`	low	4096	1843 ms	2712 ms	2060 ms	0.0%
#3	`openai/gpt-4.1-mini`	provider default	768	2416 ms	5716 ms	3241 ms	0.0%
#4	`gemini-3.7-flash`	low	4096	3278 ms	3655 ms	3372 ms	0.0%
#5	`moonshotai/kimi-k2-0905`	provider default	768	2950 ms	4790 ms	3410 ms	0.0%
#6	`google/gemini-3.5-flash-lite`	low	2048	3597 ms	4007 ms	3700 ms	0.0%
#7	`x-ai/grok-4.3`	provider default	768	942 ms	12485 ms	3828 ms	0.0%
#8	`x-ai/grok-4.3:nitro`	provider default	768	2022 ms	9654 ms	3930 ms	0.0%
#9	`google/gemini-3.6-flash`	low	768	4548 ms	4967 ms	4653 ms	0.0%
#10	`openai/gpt-5.6-luna`	none	768	4870 ms	6175 ms	5196 ms	0.0%
#11	`openai/gpt-5.6-luna`	high	4096	5496 ms	6365 ms	5713 ms	0.0%
#12	`z-ai/glm-5.2`	provider default	768	5896 ms	8454 ms	6536 ms	0.0%
#13	`google/gemini-3.7-flash`	medium	4096	6722 ms	7402 ms	6892 ms	0.0%
#14	`gemini-3.7-flash`	medium	4096	7145 ms	7336 ms	7193 ms	0.0%
#15	`google/gemini-3.7-flash`	high	4096	7992 ms	8514 ms	8123 ms	0.0%
#16	`gemini-3.7-flash`	high	4096	8506 ms	8663 ms	8545 ms	0.0%
#17	`qwen/qwen2.5-vl-72b-instruct`	provider default	768	6361 ms	20904 ms	9997 ms	0.0%
#18	`openai/gpt-5.6-sol`	none	768	11374 ms	13470 ms	11898 ms	0.0%
#19	`deepseek/deepseek-v4-flash-0731`	medium	4096	11906 ms	14207 ms	12481 ms	0.0%
#20	`z-ai/glm-4.7`	off	768	2593 ms	48244 ms	14006 ms	0.0%
#21	`deepseek/deepseek-v4-flash-0731`	max	4096	99967 ms	104620 ms	101130 ms	0.0%

Production screening

Select a run to inspect the complete evidence trail

Date	Model	Status	Preflight	Guards	Quality	Speed rank	Warm TTFT p95	Warm completion p95	Decision
2026-08-14	`gemini-3.7-flash`	qualified	12/12100.0%	8.3%	4.33	#14	7145 ms	7336 ms	quality passed; quality rank #10 of 13; speed rank #14 of 21	Inspect Run`2026-08-14/gemini-3.7-flash-google-native-medium-reasoning` qualified Decision quality passed; quality rank #10 of 13; speed rank #14 of 21 Stage judgment Candidate `gemini-3.7-flash@https://generativelanguage.googleapis.com/v1beta/openai#env:GOOGLE_API_KEY` Request profile Reasoning effort medium JSON mode on Temperature 0.7 Max tokens 4096 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 1 (8.3%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-14/gemini-3.7-flash-google-native-medium-reasoning-preflight.json`d89b5ac1ebc6f9c544f190089a8833ed3497df37a90f7b68406b637e0448f69b` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #14 of 21 Speed index 7193 ms Cold TTFT p95 5488 ms Warm TTFT p95 7145 ms Cold completion p95 5602 ms Warm completion p95 7336 ms Throughput p50 0.0 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-14/gemini-3.7-flash-google-native-medium-reasoning-perf.json`c7d2c126f1627ef04135389d9dd035e6d8f16df566b7f009c115a67edabce8e8` Quality judgment Consensus / rank 4.33 · #10 of 13 Verdict qualified Independent votes 2 pass · 0 fail · 2/2 present Self-family excluded 0 Judge spread 0.15 Usable outputs 18/18 Panel cost $1.3965 openai-sol-high`4.25 · pass`hash retained anthropic-sonnet-low`4.40 · pass`hash retained Review 67 individual API calls 12 preflight · 19 performance · 36 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-14__gemini-3.7-flash-google-native-medium-reasoning.json`0d30eb6775d3a43c6f7242626b8821332d6179eca5ffcfbb6b6e7f880bb8881a`
2026-08-14	`gemini-3.7-flash`	qualified	12/12100.0%	0.0%	4.34	#4	3278 ms	3655 ms	quality passed; quality rank #8 of 13; speed rank #4 of 21	Inspect Run`2026-08-14/gemini-3.7-flash-google-native-low-reasoning-attempt-2` qualified Decision quality passed; quality rank #8 of 13; speed rank #4 of 21 Stage judgment Candidate `gemini-3.7-flash@https://generativelanguage.googleapis.com/v1beta/openai#env:GOOGLE_API_KEY` Request profile Reasoning effort low JSON mode on Temperature 0.7 Max tokens 4096 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-14/gemini-3.7-flash-google-native-low-reasoning-attempt-2-preflight.json`c22f8e444f08fe6a2ecbcb589544cd5cb023085e3755bb74e97ab2c1f6937744` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #4 of 21 Speed index 3372 ms Cold TTFT p95 9777 ms Warm TTFT p95 3278 ms Cold completion p95 10064 ms Warm completion p95 3655 ms Throughput p50 0.0 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-14/gemini-3.7-flash-google-native-low-reasoning-attempt-2-perf.json`575e2b4750fa704cb10e76e33393d700a2863e877ad99eaaaffe97c15e182692` Quality judgment Consensus / rank 4.34 · #8 of 13 Verdict qualified Independent votes 2 pass · 0 fail · 2/2 present Self-family excluded 0 Judge spread 0.06 Usable outputs 18/18 Panel cost $1.3949 openai-sol-high`4.31 · pass`hash retained anthropic-sonnet-low`4.37 · pass`hash retained Review 67 individual API calls 12 preflight · 19 performance · 36 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-14__gemini-3.7-flash-google-native-low-reasoning-attempt-2.json`f784db4a38a4c41bdd3f9eb9caeb753c5b16169a984b14a18a28ce4b2c307d58`
2026-08-14	`gemini-3.7-flash`	qualified	12/12100.0%	0.0%	4.54	#16	8506 ms	8663 ms	quality passed; quality rank #2 of 13; speed rank #16 of 21	Inspect Run`2026-08-14/gemini-3.7-flash-google-native-high-reasoning` qualified Decision quality passed; quality rank #2 of 13; speed rank #16 of 21 Stage judgment Candidate `gemini-3.7-flash@https://generativelanguage.googleapis.com/v1beta/openai#env:GOOGLE_API_KEY` Request profile Reasoning effort high JSON mode on Temperature 0.7 Max tokens 4096 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-14/gemini-3.7-flash-google-native-high-reasoning-preflight.json`f69be041d493dad86257b3f2a43f587ec9da3486c28f1edf5e0ea50bd80cf704` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #16 of 21 Speed index 8545 ms Cold TTFT p95 6527 ms Warm TTFT p95 8506 ms Cold completion p95 6688 ms Warm completion p95 8663 ms Throughput p50 0.0 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-14/gemini-3.7-flash-google-native-high-reasoning-perf.json`9e193d0388efad4c0f6cb51650a23d3ab505574b08175628cacb5d5c43ccf02a` Quality judgment Consensus / rank 4.54 · #2 of 13 Verdict qualified Independent votes 2 pass · 0 fail · 2/2 present Self-family excluded 0 Judge spread 0.24 Usable outputs 18/18 Panel cost $1.4067 openai-sol-high`4.66 · pass`hash retained anthropic-sonnet-low`4.42 · pass`hash retained Review 67 individual API calls 12 preflight · 19 performance · 36 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-14__gemini-3.7-flash-google-native-high-reasoning.json`956a1cd29328cedb1a2c56a5fc8128eb4fdd3dcee3e5ff7488b7deb0a27aafee`
2026-08-13	`google/gemini-3.7-flash`	qualified	12/12100.0%	0.0%	4.33	#13	6722 ms	7402 ms	quality passed; quality rank #11 of 13; speed rank #13 of 21	Inspect Run`2026-08-13/gemini-3.7-flash-medium-reasoning` qualified Decision quality passed; quality rank #11 of 13; speed rank #13 of 21 Stage judgment Candidate `google/gemini-3.7-flash@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort medium JSON mode on Temperature 0.7 Max tokens 4096 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-13/gemini-3.7-flash-medium-reasoning-preflight.json`591efc5ce06871bd20e94369df45081c1fc79f206ae0cfedbc580b71697b8db5` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #13 of 21 Speed index 6892 ms Cold TTFT p95 5242 ms Warm TTFT p95 6722 ms Cold completion p95 5598 ms Warm completion p95 7402 ms Throughput p50 1218.6 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-13/gemini-3.7-flash-medium-reasoning-perf.json`f58a6de6172fb16d9239efc554d2d0de74502ab1441f1b678a43d98b24a9a5d8` Quality judgment Consensus / rank 4.33 · #11 of 13 Verdict qualified Independent votes 2 pass · 0 fail · 2/2 present Self-family excluded 0 Judge spread 0.05 Usable outputs 18/18 Panel cost $1.4160 openai-sol-high`4.30 · pass`hash retained anthropic-sonnet-low`4.35 · pass`hash retained Review 67 individual API calls 12 preflight · 19 performance · 36 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-13__gemini-3.7-flash-medium-reasoning.json`4a336547b388582ac59b17e46ac147ad0818d26523875d563407f1bfc1458fdf`
2026-08-13	`google/gemini-3.7-flash`	qualified	12/12100.0%	8.3%	4.42	#2	1843 ms	2712 ms	quality passed; quality rank #6 of 13; speed rank #2 of 21	Inspect Run`2026-08-13/gemini-3.7-flash-low-reasoning` qualified Decision quality passed; quality rank #6 of 13; speed rank #2 of 21 Stage judgment Candidate `google/gemini-3.7-flash@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort low JSON mode on Temperature 0.7 Max tokens 4096 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 1 (8.3%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-13/gemini-3.7-flash-low-reasoning-preflight.json`93e52754ea2908a5467de7a77482b2adbd98d692f3a6466db95845452931ebc0` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #2 of 21 Speed index 2060 ms Cold TTFT p95 1869 ms Warm TTFT p95 1843 ms Cold completion p95 2444 ms Warm completion p95 2712 ms Throughput p50 169.1 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-13/gemini-3.7-flash-low-reasoning-perf.json`924488549d522673f9bd4e3343dfd7724fe6902f9ec6b2edecadf887fdb6fce5` Quality judgment Consensus / rank 4.42 · #6 of 13 Verdict qualified Independent votes 2 pass · 0 fail · 2/2 present Self-family excluded 0 Judge spread 0.00 Usable outputs 18/18 Panel cost $1.4653 openai-sol-high`4.41 · pass`hash retained anthropic-sonnet-low`4.42 · pass`hash retained Review 67 individual API calls 12 preflight · 19 performance · 36 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-13__gemini-3.7-flash-low-reasoning.json`8973a0c6abcff803c0a20e8c4bde7a183b75c4d5ad05e66e8537a87088b674de`
2026-08-13	`google/gemini-3.7-flash`	qualified	12/12100.0%	0.0%	4.44	#15	7992 ms	8514 ms	quality passed; quality rank #4 of 13; speed rank #15 of 21	Inspect Run`2026-08-13/gemini-3.7-flash-high-reasoning` qualified Decision quality passed; quality rank #4 of 13; speed rank #15 of 21 Stage judgment Candidate `google/gemini-3.7-flash@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort high JSON mode on Temperature 0.7 Max tokens 4096 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-13/gemini-3.7-flash-high-reasoning-preflight.json`1ec45d752a40cb71326ecb8b41e7e751e3103eb0ad3fedd01e1c1af72a062074` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #15 of 21 Speed index 8123 ms Cold TTFT p95 7103 ms Warm TTFT p95 7992 ms Cold completion p95 7710 ms Warm completion p95 8514 ms Throughput p50 1826.3 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-13/gemini-3.7-flash-high-reasoning-perf.json`09f256c23fa24e21d0347c3c692e4ef16140fcdf3b68f20dbd2870e2c6abc8ae` Quality judgment Consensus / rank 4.44 · #4 of 13 Verdict qualified Independent votes 2 pass · 0 fail · 2/2 present Self-family excluded 0 Judge spread 0.10 Usable outputs 18/18 Panel cost $1.4633 openai-sol-high`4.49 · pass`hash retained anthropic-sonnet-low`4.39 · pass`hash retained Review 67 individual API calls 12 preflight · 19 performance · 36 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-13__gemini-3.7-flash-high-reasoning.json`ecfa2158e37451d8d7f86dfe49d95a881dbc86de982ea04fbc4e86a80d05d826`
2026-08-08	`moonshotai/kimi-k2.5:nitro`	stopped	2/2100.0%	0.0%	—	—	—	—	stopped after 2/12 calls	Inspect Run`2026-08-08/kimi-k2.5-nitro` stopped Decision stopped after 2/12 calls Stage preflight Candidate `moonshotai/kimi-k2.5:nitro@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort not captured JSON mode not captured Temperature — Max tokens — Frequency penalty — Preflight evidence Valid responses 2/2 (100.0%) Guard interventions 0 (0.0%) Observed elapsed 26282 ms–36346 ms Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/kimi-k2.5-nitro-preflight.partial.jsonl`70f21dff7ba52c7c3fc0b61b55c52b74c8fa44996ad6bbc89416699594abb2df` Performance evidence Not run: this profile did not clear preflight. Quality judgment Not judged: this profile did not clear deterministic screening. Review 2 individual API calls 2 preflight · 0 performance · 0 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__kimi-k2.5-nitro.json`c84de2cbdd02066228c5874293ca9dcdc818c9c6f96a46eeb3d8fd5d2431a518`
2026-08-08	`deepseek/deepseek-v4-pro`	rejected	11/11100.0%	18.2%	—	—	—	—	guard intervention rate (early stop)	Inspect Run`2026-08-08/deepseek-v4-pro-no-reasoning` rejected Decision guard intervention rate (early stop) Stage preflight Candidate `deepseek/deepseek-v4-pro@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort not captured JSON mode not captured Temperature — Max tokens — Frequency penalty — Preflight evidence Valid responses 11/11 (100.0%) Guard interventions 2 (18.2%) Observed elapsed 2569 ms–4773 ms Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/deepseek-v4-pro-no-reasoning-preflight.partial.jsonl`8bcd7ceac1cf4397ee188a30ac5f3b8aa6b09c1d56d1b6cb6fdcca34bffda668` Performance evidence Not run: this profile did not clear preflight. Quality judgment Not judged: this profile did not clear deterministic screening. Review 11 individual API calls 11 preflight · 0 performance · 0 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__deepseek-v4-pro-no-reasoning.json`e3a81702bfd4e874c40a6f64e8d57a38dfc59774b8420e6c21d644b4332c0fde`
2026-08-08	`deepseek/deepseek-v4-flash-0731`	rejected	1/250.0%	0.0%	—	—	—	—	structural reliability (early stop)	Inspect Run`2026-08-08/deepseek-v4-flash-0731-max-reasoning` rejected Decision structural reliability (early stop) Stage preflight Candidate `deepseek/deepseek-v4-flash-0731@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort not captured JSON mode not captured Temperature — Max tokens — Frequency penalty — Preflight evidence Valid responses 1/2 (50.0%) Guard interventions 0 (0.0%) Observed elapsed 10407 ms–15348 ms Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/deepseek-v4-flash-0731-max-reasoning-preflight.partial.jsonl`4aa848ed336de905669b57fb7cdcf73fc5ff57b4c77f53af933121278b9641e7` Performance evidence Not run: this profile did not clear preflight. Quality judgment Not judged: this profile did not clear deterministic screening. Review 2 individual API calls 2 preflight · 0 performance · 0 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__deepseek-v4-flash-0731-max-reasoning.json`d9dd9047723163a9c88b589144ee4d40b4d97caf4552c4275244518087f73aa7`
2026-08-08	`moonshotai/kimi-k2.5:nitro`	rejected	12/12100.0%	16.7%	—	—	—	—	guard intervention rate	Inspect Run`2026-08-08/kimi-k2.5-nitro-no-reasoning` rejected Decision guard intervention rate Stage preflight Candidate `moonshotai/kimi-k2.5:nitro@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort off JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 2 (16.7%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/kimi-k2.5-nitro-no-reasoning-preflight.json`5c6715172e73285a83e5df6fafc4b9484c5b56d599b9f31b46296ecceda9d1d4` Performance evidence Not run: this profile did not clear preflight. Quality judgment Not judged: this profile did not clear deterministic screening. Review 12 individual API calls 12 preflight · 0 performance · 0 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__kimi-k2.5-nitro-no-reasoning.json`9ccbba330c58163eb1389af7956f8d064bb6a73fe5bc9d91e0f425c5371908e0`
2026-08-08	`moonshotai/kimi-k2-0905:nitro`	rejected	12/12100.0%	16.7%	—	—	—	—	guard intervention rate	Inspect Run`2026-08-08/kimi-k2-0905-nitro-funded` rejected Decision guard intervention rate Stage preflight Candidate `moonshotai/kimi-k2-0905:nitro@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort provider default JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 2 (16.7%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/kimi-k2-0905-nitro-funded-preflight.json`afc70bb4906009023505b50eea72f5f924be7dd8343b51888fa3b2af5800da55` Performance evidence Not run: this profile did not clear preflight. Quality judgment Not judged: this profile did not clear deterministic screening. Review 12 individual API calls 12 preflight · 0 performance · 0 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__kimi-k2-0905-nitro-funded.json`bee63fd3ffdb1f1a35616bcfa041453d8e5f7ecec3b3e2fef47140f4cb889abf`
2026-08-08	`openai/gpt-5.6-sol`	needs judgment	12/12100.0%	0.0%	—	#18	11374 ms	13470 ms	awaiting independent judges (0/2); cloud speed rank #18 of 21	Inspect Run`2026-08-08/gpt-5.6-sol-no-reasoning` needs judgment Decision awaiting independent judges (0/2); cloud speed rank #18 of 21 Stage judgment Candidate `openai/gpt-5.6-sol@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort none JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/gpt-5.6-sol-no-reasoning-preflight.json`9c10c160528b5c49831e088346300cc5602db563dadcf5b5e9a9aed333905a5a` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #18 of 21 Speed index 11898 ms Cold TTFT p95 8907 ms Warm TTFT p95 11374 ms Cold completion p95 10159 ms Warm completion p95 13470 ms Throughput p50 137.1 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/gpt-5.6-sol-no-reasoning-perf.json`043e6eae64d2caab3d4cf4f20f9d98e69dcaffb249e1d585867e557fca302b87` Quality judgment Consensus / rank pending · unranked Verdict panel incomplete Independent votes 0 pass · 0 fail · 0/2 present Self-family excluded 1 Judge spread — Usable outputs 0/0 Panel cost $0.8796 openai-sol-high · excluded`4.17 · pass`hash retained Review 49 individual API calls 12 preflight · 19 performance · 18 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__gpt-5.6-sol-no-reasoning.json`14a335bfc3d3277052b4c1ff775e0520919b190e1680bf1b33b1cc66cbc6da1c`
2026-08-08	`openai/gpt-5.6-luna`	quality rejected	12/12100.0%	8.3%	4.43	#10	4870 ms	6175 ms	quality screen failed; quality rank #5 of 13; speed rank #10 of 21	Inspect Run`2026-08-08/gpt-5.6-luna-no-reasoning` quality rejected Decision quality screen failed; quality rank #5 of 13; speed rank #10 of 21 Stage judgment Candidate `openai/gpt-5.6-luna@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort none JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 1 (8.3%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/gpt-5.6-luna-no-reasoning-preflight.json`1cc2f609c8ff59b5b8b62af75bf8b85ff51efda5e5dfdabb41450a66a9cf973d` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #10 of 21 Speed index 5196 ms Cold TTFT p95 4105 ms Warm TTFT p95 4870 ms Cold completion p95 4618 ms Warm completion p95 6175 ms Throughput p50 526.6 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/gpt-5.6-luna-no-reasoning-perf.json`4a8aa87147b548ceb04c3578d5892b841944ecc7f060033eb31d60f7bf06a0b1` Quality judgment Consensus / rank 4.43 · #5 of 13 Verdict rejected Independent votes 0 pass · 2 fail · 2/2 present Self-family excluded 1 Judge spread 0.56 Usable outputs 17/18 Panel cost $1.7976 openai-sol-high · excluded`4.17 · pass`hash retained anthropic-sonnet-low`4.14 · fail`hash retained google-pro-high`4.71 · fail`hash retained Review 85 individual API calls 12 preflight · 19 performance · 54 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__gpt-5.6-luna-no-reasoning.json`67c98ca2023eb396c01d35e92a9ed2f854e3ab0d7c46d43c6bb67f148ab51d47`
2026-08-08	`openai/gpt-5.6-luna`	rejected	12/12100.0%	16.7%	—	—	—	—	guard intervention rate	Inspect Run`2026-08-08/gpt-5.6-luna-max-reasoning-4096` rejected Decision guard intervention rate Stage preflight Candidate `openai/gpt-5.6-luna@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort max JSON mode on Temperature 0.7 Max tokens 4096 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 2 (16.7%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/gpt-5.6-luna-max-reasoning-4096-preflight.json`a41ce3d9172d9192be9528da688b9d001ae40d0e477fbf988e7fced55aa356af` Performance evidence Not run: this profile did not clear preflight. Quality judgment Not judged: this profile did not clear deterministic screening. Review 12 individual API calls 12 preflight · 0 performance · 0 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__gpt-5.6-luna-max-reasoning-4096.json`e2186aa21dd5ca4ecf17ecb03a11af73131f91e93d5177812f1b3102a02b9fa7`
2026-08-08	`openai/gpt-5.6-luna`	qualified	12/12100.0%	0.0%	4.56	#11	5496 ms	6365 ms	quality passed; quality rank #1 of 13; speed rank #11 of 21	Inspect Run`2026-08-08/gpt-5.6-luna-high-reasoning-4096` qualified Decision quality passed; quality rank #1 of 13; speed rank #11 of 21 Stage judgment Candidate `openai/gpt-5.6-luna@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort high JSON mode on Temperature 0.7 Max tokens 4096 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/gpt-5.6-luna-high-reasoning-4096-preflight.json`a1f9cd53066eb20e5bf999333cfcad766356b87c5834e6432bd74c401487d842` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #11 of 21 Speed index 5713 ms Cold TTFT p95 6884 ms Warm TTFT p95 5496 ms Cold completion p95 7844 ms Warm completion p95 6365 ms Throughput p50 1057.1 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/gpt-5.6-luna-high-reasoning-4096-perf.json`4f81be963b60621d93fa771665f81a82f4a5465d8e2b7fb5dd0341a7c113c3cf` Quality judgment Consensus / rank 4.56 · #1 of 13 Verdict qualified Independent votes 2 pass · 0 fail · 2/2 present Self-family excluded 0 Judge spread 0.59 Usable outputs 18/18 Panel cost $0.9085 anthropic-sonnet-low`4.27 · pass`hash retained google-pro-high`4.86 · pass`hash retained Review 67 individual API calls 12 preflight · 19 performance · 36 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__gpt-5.6-luna-high-reasoning-4096.json`cc542987234778fa757304a230b3903dc8eb129e09381e611adbec3239998bd7`
2026-08-08	`z-ai/glm-4.7`	needs judgment	12/12100.0%	0.0%	3.69	#20	2593 ms	48244 ms	awaiting independent judges (1/2); cloud speed rank #20 of 21	Inspect Run`2026-08-08/glm-4.7-no-reasoning` needs judgment Decision awaiting independent judges (1/2); cloud speed rank #20 of 21 Stage judgment Candidate `z-ai/glm-4.7@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort off JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/glm-4.7-no-reasoning-preflight.json`21b0e0092a456e190a2319b243b0c83f6c273edbc3176ff1ea1ec782c01f3eb6` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #20 of 21 Speed index 14006 ms Cold TTFT p95 30590 ms Warm TTFT p95 2593 ms Cold completion p95 48121 ms Warm completion p95 48244 ms Throughput p50 52.6 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/glm-4.7-no-reasoning-perf.json`6379e39c9c8228a161d15a39b5e38f762810a2b31ffdecf3595a794a3c2c8415` Quality judgment Consensus / rank 3.69 · unranked Verdict panel incomplete Independent votes 0 pass · 1 fail · 1/2 present Self-family excluded 0 Judge spread — Usable outputs 10/18 Panel cost $0.9085 openai-sol-high`3.69 · fail`hash retained Review 49 individual API calls 12 preflight · 19 performance · 18 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__glm-4.7-no-reasoning.json`5fe07183b35014638fa26b47a45b1e71015940ed5e3f0fa4789f298710b21a42`
2026-08-08	`google/gemini-3.6-flash`	qualified	12/12100.0%	8.3%	4.45	#9	4548 ms	4967 ms	quality passed; quality rank #3 of 13; speed rank #9 of 21	Inspect Run`2026-08-08/gemini-3.6-flash-low-reasoning` qualified Decision quality passed; quality rank #3 of 13; speed rank #9 of 21 Stage judgment Candidate `google/gemini-3.6-flash@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort low JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 1 (8.3%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/gemini-3.6-flash-low-reasoning-preflight.json`13cf0a3b8e158456ccc29ee421a4182db64097822e4819a5cb35008ddb898851` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #9 of 21 Speed index 4653 ms Cold TTFT p95 2834 ms Warm TTFT p95 4548 ms Cold completion p95 3368 ms Warm completion p95 4967 ms Throughput p50 724.4 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/gemini-3.6-flash-low-reasoning-perf.json`91d1857058339851ecc5791595b83da57e8f515b36ded953b689a06b06ff23a8` Quality judgment Consensus / rank 4.45 · #3 of 13 Verdict qualified Independent votes 2 pass · 0 fail · 2/2 present Self-family excluded 0 Judge spread 0.01 Usable outputs 18/18 Panel cost $1.4074 openai-sol-high`4.44 · pass`hash retained anthropic-sonnet-low`4.45 · pass`hash retained Review 67 individual API calls 12 preflight · 19 performance · 36 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__gemini-3.6-flash-low-reasoning.json`2b27236ee3f8b482565893cdd643f1c67de1f6dead87da904ab7463be7185fd4`
2026-08-08	`google/gemini-3.5-flash-lite`	invalid profile	10/1283.3%	0.0%	—	—	—	—	superseded: insufficient completion budget	Inspect Run`2026-08-08/gemini-3.5-flash-lite-low-reasoning` invalid profile Decision superseded: insufficient completion budget Stage configuration Candidate `google/gemini-3.5-flash-lite@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort low JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 10/12 (83.3%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/gemini-3.5-flash-lite-low-reasoning-preflight.json`26f3bfeb1d5577aedd0c3e814e0cea64b4fec9e9940ffd55627e3a4fff81d539` Performance evidence Not run: this profile did not clear preflight. Quality judgment Not judged: this profile did not clear deterministic screening. Review 14 individual API calls 12 preflight · 0 performance · 0 judgment · 2 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__gemini-3.5-flash-lite-low-reasoning.json`b2a7f94345e687c4d3d76cb7ebfac35e9f6c1620d224723230208e97c37193ef`
2026-08-08	`google/gemini-3.5-flash-lite`	qualified	12/12100.0%	0.0%	4.41	#6	3597 ms	4007 ms	quality passed; quality rank #7 of 13; speed rank #6 of 21	Inspect Run`2026-08-08/gemini-3.5-flash-lite-low-reasoning-2048` qualified Decision quality passed; quality rank #7 of 13; speed rank #6 of 21 Stage judgment Candidate `google/gemini-3.5-flash-lite@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort low JSON mode on Temperature 0.7 Max tokens 2048 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/gemini-3.5-flash-lite-low-reasoning-2048-preflight.json`b6478393bfc69ff9c97a1605c9cb5d7733d96170bc3807c69b486120a7551161` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #6 of 21 Speed index 3700 ms Cold TTFT p95 3641 ms Warm TTFT p95 3597 ms Cold completion p95 3751 ms Warm completion p95 4007 ms Throughput p50 1574.5 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/gemini-3.5-flash-lite-low-reasoning-2048-perf.json`7a028b4da4f76e403882cf17d71329a6f4cfd23ce6fad642bb34f2993ed117ac` Quality judgment Consensus / rank 4.41 · #7 of 13 Verdict qualified Independent votes 2 pass · 0 fail · 2/2 present Self-family excluded 0 Judge spread 0.07 Usable outputs 18/18 Panel cost $1.4522 openai-sol-high`4.44 · pass`hash retained anthropic-sonnet-low`4.37 · pass`hash retained Review 67 individual API calls 12 preflight · 19 performance · 36 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__gemini-3.5-flash-lite-low-reasoning-2048.json`5d6963f2c4d34cc66393f9cfcce4dccab7da4bff3185ecb2bc8ba99a8337c4e4`
2026-08-08	`deepseek-v4-flash`	rejected	12/12100.0%	16.7%	—	—	—	—	guard intervention rate	Inspect Run`2026-08-08/deepseek-v4-flash-direct-high-reasoning-4096` rejected Decision guard intervention rate Stage preflight Candidate `deepseek-v4-flash@https://api.deepseek.com/v1#env:DEEPSEEK_API_KEY` Request profile Reasoning effort high JSON mode on Temperature 0.7 Max tokens 4096 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 2 (16.7%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/deepseek-v4-flash-direct-high-reasoning-4096-preflight.json`6163d9e14fbc564dc2ee59ac6054da0f7971b13e23fefc0ad7960b46c6737f4e` Performance evidence Not run: this profile did not clear preflight. Quality judgment Not judged: this profile did not clear deterministic screening. Review 12 individual API calls 12 preflight · 0 performance · 0 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__deepseek-v4-flash-direct-high-reasoning-4096.json`dc7adbc0596f81e4e18e99c1a0849e2cdfa88a688a051263c329653ae95a2631`
2026-08-08	`deepseek/deepseek-v4-flash-0731`	rejected	12/12100.0%	25.0%	—	—	—	—	guard intervention rate	Inspect Run`2026-08-08/deepseek-v4-flash-0731-no-reasoning` rejected Decision guard intervention rate Stage preflight Candidate `deepseek/deepseek-v4-flash-0731@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort off JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 3 (25.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/deepseek-v4-flash-0731-no-reasoning-preflight.json`e3574e4f1f343ed74b2f09198bb2ba46f88840d3b3fe7e00ded4f4c6e60306ac` Performance evidence Not run: this profile did not clear preflight. Quality judgment Not judged: this profile did not clear deterministic screening. Review 12 individual API calls 12 preflight · 0 performance · 0 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__deepseek-v4-flash-0731-no-reasoning.json`d00dc324c47749fd7570a8a43423125d2d7775c6ed0f33dbae322c4bac8a501c`
2026-08-08	`deepseek/deepseek-v4-flash-0731`	needs judgment	12/12100.0%	0.0%	4.37	#19	11906 ms	14207 ms	awaiting independent judges (1/2); cloud speed rank #19 of 21	Inspect Run`2026-08-08/deepseek-v4-flash-0731-medium-reasoning-4096` needs judgment Decision awaiting independent judges (1/2); cloud speed rank #19 of 21 Stage judgment Candidate `deepseek/deepseek-v4-flash-0731@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort medium JSON mode on Temperature 0.7 Max tokens 4096 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/deepseek-v4-flash-0731-medium-reasoning-4096-preflight.json`6330fe5aa8ed64be084628c86679f0124b0eecc7fcf4cbf846a82e2462a748cf` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #19 of 21 Speed index 12481 ms Cold TTFT p95 29019 ms Warm TTFT p95 11906 ms Cold completion p95 31051 ms Warm completion p95 14207 ms Throughput p50 202.5 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/deepseek-v4-flash-0731-medium-reasoning-4096-perf.json`eb911f619608a3bea35a1e84169901def20cba8e37c1388f568d33daa969f2cf` Quality judgment Consensus / rank 4.37 · unranked Verdict panel incomplete Independent votes 0 pass · 1 fail · 1/2 present Self-family excluded 0 Judge spread — Usable outputs 17/18 Panel cost $0.5363 anthropic-sonnet-low`4.37 · fail`hash retained Review 49 individual API calls 12 preflight · 19 performance · 18 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__deepseek-v4-flash-0731-medium-reasoning-4096.json`1c4db6c4ce5ac77b85dab78878e7a1b2644735879f43974d0899bac7a7410fd5`
2026-08-08	`deepseek/deepseek-v4-flash-0731`	needs adjudication	12/12100.0%	0.0%	4.50	#21	99967 ms	104620 ms	judge disagreement requires adjudication; speed rank #21 of 21	Inspect Run`2026-08-08/deepseek-v4-flash-0731-max-reasoning-4096` needs adjudication Decision judge disagreement requires adjudication; speed rank #21 of 21 Stage judgment Candidate `deepseek/deepseek-v4-flash-0731@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort max JSON mode on Temperature 0.7 Max tokens 4096 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/deepseek-v4-flash-0731-max-reasoning-4096-preflight.json`1d822e6cd8d55981961aee05612cbbc37ae4f44728a37cc4aa3b2c851c9cbbc2` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #21 of 21 Speed index 101130 ms Cold TTFT p95 123127 ms Warm TTFT p95 99967 ms Cold completion p95 124790 ms Warm completion p95 104620 ms Throughput p50 409.1 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/deepseek-v4-flash-0731-max-reasoning-4096-perf.json`cbb65974797d08374a2dbaa02a78954a96e57be961d53c0250964b740cda6f7a` Quality judgment Consensus / rank 4.50 · unranked Verdict adjudication required Independent votes 1 pass · 1 fail · 2/2 present Self-family excluded 0 Judge spread 0.43 Usable outputs 18/18 Panel cost $0.8898 anthropic-sonnet-low`4.28 · fail`hash retained google-pro-high`4.72 · pass`hash retained Review 67 individual API calls 12 preflight · 19 performance · 36 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__deepseek-v4-flash-0731-max-reasoning-4096.json`2d596d86d95e89ff7532cdac29a0e04bbcf3eb54729e1ab999b4938715c88e43`
2026-08-08	`deepseek/deepseek-v4-flash-0731`	rejected	10/1283.3%	25.0%	—	—	—	—	structural reliability	Inspect Run`2026-08-08/deepseek-v4-flash-0731-low-reasoning-4096` rejected Decision structural reliability Stage preflight Candidate `deepseek/deepseek-v4-flash-0731@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort low JSON mode on Temperature 0.7 Max tokens 4096 Frequency penalty 0.5 Preflight evidence Valid responses 10/12 (83.3%) Guard interventions 3 (25.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/deepseek-v4-flash-0731-low-reasoning-4096-preflight.json`d50acf3319f44964a578aed95bacd04ac401388c6b935b2cf4005f6e0f08dbde` Performance evidence Not run: this profile did not clear preflight. Quality judgment Not judged: this profile did not clear deterministic screening. Review 12 individual API calls 12 preflight · 0 performance · 0 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__deepseek-v4-flash-0731-low-reasoning-4096.json`999da849729e787c352683c92fefdae9926d9102d3c2d834b2f2f89c1bf3a05e`
2026-08-08	`deepseek/deepseek-v4-flash-0731`	rejected	12/12100.0%	16.7%	—	—	—	—	guard intervention rate	Inspect Run`2026-08-08/deepseek-v4-flash-0731-high-reasoning-4096` rejected Decision guard intervention rate Stage preflight Candidate `deepseek/deepseek-v4-flash-0731@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort high JSON mode on Temperature 0.7 Max tokens 4096 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 2 (16.7%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-08-08/deepseek-v4-flash-0731-high-reasoning-4096-preflight.json`f1bc20c3089aa71859ad840a47ee7344e359a2187ab61994412eea0c3c252764` Performance evidence Not run: this profile did not clear preflight. Quality judgment Not judged: this profile did not clear deterministic screening. Review 12 individual API calls 12 preflight · 0 performance · 0 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-08-08__deepseek-v4-flash-0731-high-reasoning-4096.json`60aba7a21f2e42728d0e959d4340fa9fd5ab096919995643f6dd05f613a0d0c2`
2026-07-27	`qwen/qwen2.5-vl-72b-instruct`	quality rejected	12/12100.0%	0.0%	3.71	#17	6361 ms	20904 ms	quality screen failed; quality rank #13 of 13; speed rank #17 of 21	Inspect Run`2026-07-27/qwen25-vl-72b` quality rejected Decision quality screen failed; quality rank #13 of 13; speed rank #17 of 21 Stage judgment Candidate `qwen/qwen2.5-vl-72b-instruct@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort provider default JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/qwen25-vl-72b-preflight.json`97f31fad3663428c2c5200352e9ce61941a5a06d0202c35da9c8766f1790ccf5` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #17 of 21 Speed index 9997 ms Cold TTFT p95 4494 ms Warm TTFT p95 6361 ms Cold completion p95 19988 ms Warm completion p95 20904 ms Throughput p50 8.1 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/qwen25-vl-72b-perf.json`1256dc2692f28e7c7a02913111544d287d63293eb5ccaa77c88b63c24ac2659a` Quality judgment Consensus / rank 3.71 · #13 of 13 Verdict rejected Independent votes 0 pass · 3 fail · 3/2 present Self-family excluded 0 Judge spread 0.24 Usable outputs 18/18 Panel cost $1.7151 openai-sol-high`3.71 · fail`hash retained anthropic-sonnet-low`3.64 · fail`hash retained google-pro-high`3.89 · fail`hash retained Review 85 individual API calls 12 preflight · 19 performance · 54 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-07-27__qwen25-vl-72b.json`19a996901b382e357fdcd9191d7d431220fdd0b8d49c76e6cfdba2f4e1cae62e`
2026-07-27	`nvidia/nemotron-3-ultra-550b-a55b`	rejected	4/1233.3%	25.0%	—	—	—	—	structural reliability	Inspect Run`2026-07-27/nemotron-3-ultra` rejected Decision structural reliability Stage preflight Candidate `nvidia/nemotron-3-ultra-550b-a55b@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort provider default JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 4/12 (33.3%) Guard interventions 3 (25.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/nemotron-3-ultra-preflight.json`841804932eda683370b609b05bbecda02911be129ff41bb17e808491868ec051` Performance evidence Not run: this profile did not clear preflight. Quality judgment Not judged: this profile did not clear deterministic screening. Review 12 individual API calls 12 preflight · 0 performance · 0 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-07-27__nemotron-3-ultra.json`67f2085c44cb271a33f10dddf2532fd769b9cad3e6f313b74476bf3093bcbfc1`
2026-07-27	`mistralai/mistral-medium-3.1`	rejected	12/12100.0%	41.7%	—	—	—	—	guard intervention rate	Inspect Run`2026-07-27/mistral-medium-3.1` rejected Decision guard intervention rate Stage preflight Candidate `mistralai/mistral-medium-3.1@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort provider default JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 5 (41.7%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/mistral-medium-3.1-preflight.json`2982f5f6e25603ceb85500a9c0a893f5f3a8fd20637f5a0653085a3913c7dd4c` Performance evidence Not run: this profile did not clear preflight. Quality judgment Not judged: this profile did not clear deterministic screening. Review 12 individual API calls 12 preflight · 0 performance · 0 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-07-27__mistral-medium-3.1.json`8fb2efa0e01f8a4c6f1d5bb52fc7b2550c12b66c7efa6a00c7639694296ef74d`
2026-07-27	`moonshotai/kimi-k2-0905`	qualified	12/12100.0%	0.0%	4.34	#5	2950 ms	4790 ms	quality passed; quality rank #9 of 13; speed rank #5 of 21	Inspect Run`2026-07-27/kimi-k2-0905` qualified Decision quality passed; quality rank #9 of 13; speed rank #5 of 21 Stage judgment Candidate `moonshotai/kimi-k2-0905@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort provider default JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/kimi-k2-0905-preflight.json`4c10ed7a7639e6d552e069ea3eca09f8a94349721a96c19af14f65de8c5bd61e` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #5 of 21 Speed index 3410 ms Cold TTFT p95 2912 ms Warm TTFT p95 2950 ms Cold completion p95 5510 ms Warm completion p95 4790 ms Throughput p50 40.4 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/kimi-k2-0905-perf.json`0d2cddcbf6a059beb3206c12242e0aa50e8623f6ce7770165040563717d8d55f` Quality judgment Consensus / rank 4.34 · #9 of 13 Verdict qualified Independent votes 3 pass · 0 fail · 3/2 present Self-family excluded 0 Judge spread 0.21 Usable outputs 18/18 Panel cost $1.7905 openai-sol-high`4.17 · pass`hash retained anthropic-sonnet-low`4.34 · pass`hash retained google-pro-high`4.37 · pass`hash retained Review 85 individual API calls 12 preflight · 19 performance · 54 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-07-27__kimi-k2-0905.json`e4f54d1aa51bd966e9aee380cd01d953bb1c98f33cbee544c18954b7932a362e`
2026-07-27	`moonshotai/kimi-k2-0905:nitro`	needs adjudication	12/12100.0%	0.0%	4.14	#1	1336 ms	1344 ms	judge disagreement requires adjudication; speed rank #1 of 21	Inspect Run`2026-07-27/kimi-k2-0905-nitro` needs adjudication Decision judge disagreement requires adjudication; speed rank #1 of 21 Stage judgment Candidate `moonshotai/kimi-k2-0905:nitro@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort provider default JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/kimi-k2-0905-nitro-preflight.json`a693e679b53abff4ff555220671625d39e44fd7478428318e7d465d713755104` Performance evidence Measurements 48 Cold / warm 6 / 42 Speed rank #1 of 21 Speed index 1338 ms Cold TTFT p95 1664 ms Warm TTFT p95 1336 ms Cold completion p95 1681 ms Warm completion p95 1344 ms Throughput p50 6000.0 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/kimi-k2-0905-nitro-perf-expanded.json`1984ba432a4cbb82e60be1a1fc52a2fc36828d2a2654a41229fd4e052d258133` Quality judgment Consensus / rank 4.14 · unranked Verdict adjudication required Independent votes 1 pass · 2 fail · 3/2 present Self-family excluded 0 Judge spread 0.84 Usable outputs 18/18 Panel cost $1.8702 openai-sol-high`3.92 · fail`hash retained anthropic-sonnet-low`4.14 · fail`hash retained google-pro-high`4.76 · pass`hash retained Review 115 individual API calls 12 preflight · 49 performance · 54 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-07-27__kimi-k2-0905-nitro.json`92eb221e48ab8bb8785cc86ad7a5dfa261eb3928b330456345fa3522096be7a3`
2026-07-27	`x-ai/grok-4.3`	needs adjudication	12/12100.0%	0.0%	4.12	#7	942 ms	12485 ms	judge disagreement requires adjudication; speed rank #7 of 21	Inspect Run`2026-07-27/grok-4.3` needs adjudication Decision judge disagreement requires adjudication; speed rank #7 of 21 Stage judgment Candidate `x-ai/grok-4.3@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort provider default JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/grok-4.3-preflight.json`2c86351404f17832eb303ed2516c9b91877ec04f6bb7af6807e1a7e8501f1308` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #7 of 21 Speed index 3828 ms Cold TTFT p95 1402 ms Warm TTFT p95 942 ms Cold completion p95 13085 ms Warm completion p95 12485 ms Throughput p50 87.4 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/grok-4.3-perf.json`21c39b8de9ac7094b9b363307bf4814b95296b53abce0d6afc3af6d862567d4b` Quality judgment Consensus / rank 4.12 · unranked Verdict adjudication required Independent votes 2 pass · 1 fail · 3/2 present Self-family excluded 0 Judge spread 1.08 Usable outputs 18/18 Panel cost $1.8462 openai-sol-high`3.13 · fail`hash retained anthropic-sonnet-low`4.12 · pass`hash retained google-pro-high`4.21 · pass`hash retained Review 85 individual API calls 12 preflight · 19 performance · 54 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-07-27__grok-4.3.json`bbb27a87fd7e7707ee9b67fd13574a985feab15d7009def148263d5a341e0635`
2026-07-27	`x-ai/grok-4.3:nitro`	needs adjudication	12/12100.0%	0.0%	3.94	#8	2022 ms	9654 ms	judge disagreement requires adjudication; speed rank #8 of 21	Inspect Run`2026-07-27/grok-4.3-nitro` needs adjudication Decision judge disagreement requires adjudication; speed rank #8 of 21 Stage judgment Candidate `x-ai/grok-4.3:nitro@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort provider default JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/grok-4.3-nitro-preflight.json`8ad22e9ce388b6be2ac2153583bf9578080341f1789f70b4d15dc67be320a30c` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #8 of 21 Speed index 3930 ms Cold TTFT p95 1086 ms Warm TTFT p95 2022 ms Cold completion p95 11562 ms Warm completion p95 9654 ms Throughput p50 103.9 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/grok-4.3-nitro-perf.json`28458bc124df89f5cc37175ae044b4792eb0ffec126725c6da23e8795681948a` Quality judgment Consensus / rank 3.94 · unranked Verdict adjudication required Independent votes 1 pass · 2 fail · 3/2 present Self-family excluded 0 Judge spread 0.95 Usable outputs 18/18 Panel cost $1.8816 openai-sol-high`3.33 · fail`hash retained anthropic-sonnet-low`3.94 · fail`hash retained google-pro-high`4.28 · pass`hash retained Review 85 individual API calls 12 preflight · 19 performance · 54 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-07-27__grok-4.3-nitro.json`a0311521ecdff6c4a4f12fb7ba426b31a2f260bcba7e851675684b0a372092a3`
2026-07-27	`openai/gpt-oss-120b:nitro`	rejected	9/1275.0%	0.0%	—	—	—	—	structural reliability	Inspect Run`2026-07-27/gpt-oss-120b-nitro` rejected Decision structural reliability Stage preflight Candidate `openai/gpt-oss-120b:nitro@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort provider default JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 9/12 (75.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/gpt-oss-120b-nitro-preflight.json`267c0bb8c91dfb5d892fb8966144c4a3c9f2ebfa72ca077ea5a51f08fe679325` Performance evidence Not run: this profile did not clear preflight. Quality judgment Not judged: this profile did not clear deterministic screening. Review 12 individual API calls 12 preflight · 0 performance · 0 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-07-27__gpt-oss-120b-nitro.json`1d4ad179de0984e3fd68c14b9923dc54ada4ff33a94058d2b4439debc334f884`
2026-07-27	`openai/gpt-4o`	rejected	12/12100.0%	16.7%	—	—	—	—	guard intervention rate	Inspect Run`2026-07-27/gpt-4o` rejected Decision guard intervention rate Stage preflight Candidate `openai/gpt-4o@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort provider default JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 2 (16.7%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/gpt-4o-preflight.json`bcd5308204ea1c01d3dfa6ee675f0b1908d720b28fc262090e50cb21e481acc3` Performance evidence Not run: this profile did not clear preflight. Quality judgment Not judged: this profile did not clear deterministic screening. Review 12 individual API calls 12 preflight · 0 performance · 0 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-07-27__gpt-4o.json`9f80067136a99b7baaf440d5454d2175730624399c3be5df31ab600db390bef4`
2026-07-27	`openai/gpt-4.1-mini`	quality rejected	12/12100.0%	0.0%	3.77	#3	2416 ms	5716 ms	quality screen failed; quality rank #12 of 13; speed rank #3 of 21	Inspect Run`2026-07-27/gpt-4.1-mini` quality rejected Decision quality screen failed; quality rank #12 of 13; speed rank #3 of 21 Stage judgment Candidate `openai/gpt-4.1-mini@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort provider default JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/gpt-4.1-mini-preflight.json`9e3801a1c127ebf228c59acde1edc85d88f83a207be481b04b4aa9685d05957e` Performance evidence Measurements 48 Cold / warm 6 / 42 Speed rank #3 of 21 Speed index 3241 ms Cold TTFT p95 1116 ms Warm TTFT p95 2416 ms Cold completion p95 3605 ms Warm completion p95 5716 ms Throughput p50 73.2 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/gpt-4.1-mini-perf-expanded.json`9b7a22a2670f17d09cd502984ee682e81e0997c1ceff19ff6830d16035df080b` Quality judgment Consensus / rank 3.77 · #12 of 13 Verdict rejected Independent votes 0 pass · 2 fail · 2/2 present Self-family excluded 1 Judge spread 0.72 Usable outputs 18/18 Panel cost $1.7797 openai-sol-high · excluded`3.66 · fail`hash retained anthropic-sonnet-low`3.41 · fail`hash retained google-pro-high`4.13 · fail`hash retained Review 115 individual API calls 12 preflight · 49 performance · 54 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-07-27__gpt-4.1-mini.json`2b77583cb395cb3cd30215632ac56d2e3c4ee3a341589e67e9106321fab4a5f9`
2026-07-27	`z-ai/glm-5.2`	needs adjudication	12/12100.0%	0.0%	4.18	#12	5896 ms	8454 ms	judge disagreement requires adjudication; speed rank #12 of 21	Inspect Run`2026-07-27/glm-5.2` needs adjudication Decision judge disagreement requires adjudication; speed rank #12 of 21 Stage judgment Candidate `z-ai/glm-5.2@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort provider default JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 12/12 (100.0%) Guard interventions 0 (0.0%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/glm-5.2-preflight.json`6c7095353754934c61fd6f8471d616445006c6b187f5c8ff1a64b7586ecd88c1` Performance evidence Measurements 18 Cold / warm 6 / 12 Speed rank #12 of 21 Speed index 6536 ms Cold TTFT p95 2166 ms Warm TTFT p95 5896 ms Cold completion p95 10476 ms Warm completion p95 8454 ms Throughput p50 61.3 tok/s Error rate 0.0% Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/glm-5.2-perf.json`be49dbe1898676b4c19fe866be95e8641a0f0a510053dbea16fe800e6d6fcb88` Quality judgment Consensus / rank 4.18 · unranked Verdict adjudication required Independent votes 1 pass · 2 fail · 3/2 present Self-family excluded 0 Judge spread 1.69 Usable outputs 18/18 Panel cost $1.8963 openai-sol-high`3.04 · fail`hash retained anthropic-sonnet-low`4.18 · fail`hash retained google-pro-high`4.72 · pass`hash retained Review 85 individual API calls 12 preflight · 19 performance · 54 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-07-27__glm-5.2.json`2d6225758ed4f93c2f473e96c577610bdb72794ff8eab1b329bfd91720d2ff80`
2026-07-27	`z-ai/glm-4.5v`	rejected	10/1283.3%	33.3%	—	—	—	—	structural reliability	Inspect Run`2026-07-27/glm-4.5v` rejected Decision structural reliability Stage preflight Candidate `z-ai/glm-4.5v@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort provider default JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 10/12 (83.3%) Guard interventions 4 (33.3%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/glm-4.5v-preflight.json`525e1ec5f33b05372d050ca7cddde3e7cd7d239c278f37d61cacd9487e03744f` Performance evidence Not run: this profile did not clear preflight. Quality judgment Not judged: this profile did not clear deterministic screening. Review 12 individual API calls 12 preflight · 0 performance · 0 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-07-27__glm-4.5v.json`ee29de33fedb9cd913fd48a72a30e1ff2271aaf5fb5e3ab26a62bd2a435f4f9e`
2026-07-27	`google/gemini-2.5-flash`	rejected	11/1291.7%	8.3%	—	—	—	—	structural reliability	Inspect Run`2026-07-27/gemini-2.5-flash` rejected Decision structural reliability Stage preflight Candidate `google/gemini-2.5-flash@https://openrouter.ai/api/v1#env:OPENROUTER_API_KEY` Request profile Reasoning effort provider default JSON mode on Temperature 0.7 Max tokens 768 Frequency penalty 0.5 Preflight evidence Valid responses 11/12 (91.7%) Guard interventions 1 (8.3%) Observed elapsed not captured Raw receipt archived outside Git docs/proofs/cloud-dialogue-qualification/runs/2026-07-27/gemini-2.5-flash-preflight.json`d01665f8eafd611e0f8be34dafbae53b036e5d798f85c82541f82f0469a86198` Performance evidence Not run: this profile did not clear preflight. Quality judgment Not judged: this profile did not clear deterministic screening. Review 12 individual API calls 12 preflight · 0 performance · 0 judgment · 0 diagnostic replay Open this section to load the immutable call evidence. qualification-calls/2026-07-27__gemini-2.5-flash.json`ae392e2e4ded77403a24773c171ba5528bc027d643295302dc91b6ac2dc9ed84`

Dialogue qualification funnel

Screening policy

Blinded dialogue quality ranking

Qualified cloud speed ranking

Production screening

Decision

Request profile

Preflight evidence

Performance evidence

Quality judgment

Decision

Request profile

Preflight evidence

Performance evidence

Quality judgment

Decision

Request profile

Preflight evidence

Performance evidence

Quality judgment

Decision

Request profile

Preflight evidence

Performance evidence

Quality judgment

Decision

Request profile

Preflight evidence

Performance evidence

Quality judgment

Decision

Request profile

Preflight evidence

Performance evidence

Quality judgment

Decision

Request profile

Preflight evidence

Performance evidence

Quality judgment

Decision

Request profile

Preflight evidence

Performance evidence

Quality judgment

Decision

Request profile

Preflight evidence

Performance evidence

Quality judgment

Decision

Request profile

Preflight evidence

Performance evidence

Quality judgment

Decision

Request profile

Preflight evidence

Performance evidence

Quality judgment

Decision

Request profile

Preflight evidence

Performance evidence

Quality judgment

Decision

Request profile

Preflight evidence

Performance evidence

Quality judgment

Decision

Request profile

Preflight evidence

Performance evidence

Quality judgment

Decision

Request profile

Preflight evidence

Performance evidence

Quality judgment