Runs on this setup
Earlier runs keep their own engine versions and outcomes. Workload shapes stay separate.
History and measured sweep
| Date | Shape | Each person · median | Machine total | First word · median | Result | Run |
|---|---|---|---|---|---|---|
2026-10-07 | 5 people | not recorded | not recorded | not recorded | Failed | |
2026-10-07 | 1 people | not recorded | not recorded | not recorded | Failed | |
2026-10-07 | 2 people | not recorded | not recorded | not recorded | Failed | |
2026-10-07 | 2 people | not recorded | not recorded | not recorded | Failed | |
2026-10-07 | 2 people | not recorded | not recorded | not recorded | Failed | |
2026-10-07 | 2 people | not recorded | not recorded | not recorded | Failed | |
2026-10-07 | 2 people | not recorded | not recorded | not recorded | Failed | |
2026-10-07 | 2 people | not recorded | not recorded | not recorded | Failed | |
2026-10-07 | 5 people | not recorded | not recorded | not recorded | Failed | |
2026-10-07 | 1 people | not recorded | not recorded | not recorded | Failed | |
2026-10-07 | 1 people | not recorded | not recorded | not recorded | Failed | |
2026-10-07 | 5 people | not recorded | not recorded | not recorded | Failed |
Answer checks
5/50 turns
Rolling-chat turn checks
| Person · turn | Filled input | First word | Recent · pinned · old | Result | Run |
|---|---|---|---|---|---|
1 · 1 | 20195 | 5.51 s | true · true · recalled | Passed | |
1 · 2 | 20255 | 4.01 s | true · true · incorrect | Failed | |
1 · 3 | 20256 | 4.01 s | true · true · incorrect | Failed | |
1 · 4 | 20258 | 4.05 s | true · true · incorrect | Failed | |
1 · 5 | 20254 | 4.01 s | true · true · incorrect | Failed | |
1 · 6 | 20257 | 4.03 s | true · true · incorrect | Failed | |
1 · 7 | 20255 | 4.03 s | true · true · incorrect | Failed | |
1 · 8 | 20254 | 4.04 s | true · true · incorrect | Failed | |
1 · 9 | 20255 | 4.02 s | true · true · incorrect | Failed | |
1 · 10 | 20259 | 4.02 s | true · true · incorrect | Failed | |
2 · 1 | 20201 | 3.96 s | true · true · recalled | Passed | |
2 · 2 | 20257 | 4.01 s | true · true · incorrect | Failed | |
2 · 3 | 20257 | 4.02 s | true · true · incorrect | Failed | |
2 · 4 | 20260 | 4.02 s | true · true · incorrect | Failed | |
2 · 5 | 20261 | 4.02 s | true · true · incorrect | Failed | |
2 · 6 | 20257 | 4.02 s | true · true · incorrect | Failed | |
2 · 7 | 20261 | 4.02 s | true · true · incorrect | Failed | |
2 · 8 | 20259 | 4.02 s | true · true · incorrect | Failed | |
2 · 9 | 20259 | 4.03 s | true · true · incorrect | Failed | |
2 · 10 | 20261 | 4.04 s | true · true · incorrect | Failed | |
3 · 1 | 20196 | 3.95 s | true · true · recalled | Passed | |
3 · 2 | 20257 | 4.01 s | true · true · incorrect | Failed | |
3 · 3 | 20259 | 4.02 s | true · true · incorrect | Failed | |
3 · 4 | 20256 | 4.02 s | true · true · incorrect | Failed | |
3 · 5 | 20259 | 4.02 s | true · true · incorrect | Failed | |
3 · 6 | 20256 | 4.01 s | true · true · incorrect | Failed | |
3 · 7 | 20256 | 4.02 s | true · true · incorrect | Failed | |
3 · 8 | 20256 | 4.02 s | true · true · incorrect | Failed | |
3 · 9 | 20256 | 4.01 s | true · true · incorrect | Failed | |
3 · 10 | 20261 | 4.04 s | true · true · incorrect | Failed | |
4 · 1 | 20199 | 3.95 s | true · true · recalled | Passed | |
4 · 2 | 20259 | 4.01 s | true · true · incorrect | Failed | |
4 · 3 | 20262 | 4.02 s | true · true · incorrect | Failed | |
4 · 4 | 20259 | 4.01 s | true · true · incorrect | Failed | |
4 · 5 | 20258 | 4.01 s | true · true · incorrect | Failed | |
4 · 6 | 20259 | 4.02 s | true · true · incorrect | Failed | |
4 · 7 | 20260 | 4.02 s | true · true · incorrect | Failed | |
4 · 8 | 20261 | 4.03 s | true · true · incorrect | Failed | |
4 · 9 | 20259 | 4.03 s | true · true · incorrect | Failed | |
4 · 10 | 20259 | 4.04 s | true · true · incorrect | Failed | |
5 · 1 | 20201 | 3.96 s | true · true · recalled | Passed | |
5 · 2 | 20259 | 4.02 s | true · true · incorrect | Failed | |
5 · 3 | 20260 | 4.02 s | true · true · incorrect | Failed | |
5 · 4 | 20258 | 4.04 s | true · true · incorrect | Failed | |
5 · 5 | 20258 | 4.03 s | true · true · incorrect | Failed | |
5 · 6 | 20259 | 4.04 s | true · true · incorrect | Failed | |
5 · 7 | 20259 | 4.03 s | true · true · incorrect | Failed | |
5 · 8 | 20258 | 4.04 s | true · true · incorrect | Failed | |
5 · 9 | 20259 | 4.04 s | true · true · incorrect | Failed | |
5 · 10 | 20264 | 4.05 s | true · true · incorrect | Failed |
Method
Staggered multi-turn fictional archive, exact engine-tokenized rolling gateway; independent recent, pinned and old probes each turn.
Limits
Bounded tuning pilot; incomplete window growth cannot pass the full workload grade. Lab retrieval and rolling-history evidence only. No summaries. Shared engine cache intervals include overlapping users and are not per-user hits. Gateway path, not authenticated document ingestion or customer acceptance.
What did not work
- 5/50 requests passed.
What ran
Exact identity
- Original publisher model
- Qwen3.6-35B-A3B
- Publisher revision
995ad96eacd98c81ed38be0c5b274b04031597b0- Variant · format · precision
- NVFP4 variant · safetensors · NVFP4
- Weights SHA-256
- not recorded
- Engine · version · feature
- NVIDIA NGC vLLM ·
0.29.0+5013de39.dev· plain - Hardware
- NVIDIA GB10, 128 GB unified memory
- Harness / workload
keep-rolling-context-v1 · 1- Workload suite · version
not recorded- Harness SHA-256
not recorded- Fixture SHA-256
not recorded
Workload shape
- People in this run
- 5
- Configured window (tokens)
- 262144
- Configured window (tokens)
- 262144
- Output token ceiling
- 256
Full reproduction pack available to customers and evaluators on request
All recorded numbers
Metric definitions, scope and statistics
| Metric | Value | Scope | Statistic | Sample count | Basis / source field | Run |
|---|---|---|---|---|---|---|
roll ttft max seconds | 4.05 s | per person | max | not recorded | roll_ttft_max_seconds | |
requests ok | 5.00 count | machine total | single | not recorded | requests_ok | |
roll ttft p50 seconds | 4.02 s | per person | median | not recorded | roll_ttft_p50_seconds | |
rolled off abstained | 0.00 count | machine total | single | not recorded | rolled_off_abstained | |
rolled off invalid | 0.00 count | machine total | single | not recorded | rolled_off_invalid | |
rolled off wrong | 45.00 count | machine total | single | not recorded | rolled_off_wrong | |
normal ttft max seconds | 5.51 s | per person | max | not recorded | normal_ttft_max_seconds | |
rolled off correct | 0.00 count | machine total | single | not recorded | rolled_off_correct | |
normal ttft p50 seconds | 3.96 s | per person | median | not recorded | normal_ttft_p50_seconds | |
rolled off total | 45.00 count | machine total | single | not recorded | rolled_off_total | |
requests total | 50.00 count | machine total | single | not recorded | requests_total |
Data and citation
Skilak Keep evidence, run ac1380ca0c6a4717af067bec, measured 2026-10-07T16:29:41.609051+00:00. https://skilak.ai/keep/resources/evidence/r/ac1380ca0c6a4717af067bec
