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 | Passed | |
2026-10-07 | 5 people | not recorded | not recorded | not recorded | Failed | |
2026-10-07 | 1 people | not recorded | not recorded | not recorded | Failed |
Answer checks
50/50 turns
Rolling-chat turn checks
| Person · turn | Filled input | First word | Recent · pinned · old | Result | Run |
|---|---|---|---|---|---|
1 · 1 | 20208 | 21.25 s | true · true · recalled | Passed | |
1 · 2 | 20377 | 21.54 s | true · true · recalled | Passed | |
1 · 3 | 20379 | 34.63 s | true · true · recalled | Passed | |
1 · 4 | 20381 | 27.16 s | true · true · recalled | Passed | |
1 · 5 | 20377 | 35.95 s | true · true · recalled | Passed | |
1 · 6 | 20377 | 38.84 s | true · true · recalled | Passed | |
1 · 7 | 20379 | 42.27 s | true · true · recalled | Passed | |
1 · 8 | 20378 | 61.35 s | true · true · recalled | Passed | |
1 · 9 | 20379 | 51.63 s | true · true · recalled | Passed | |
1 · 10 | 20384 | 48.37 s | true · true · recalled | Passed | |
2 · 1 | 20214 | 35.95 s | true · true · recalled | Passed | |
2 · 2 | 20380 | 28.92 s | true · true · recalled | Passed | |
2 · 3 | 20380 | 45.25 s | true · true · recalled | Passed | |
2 · 4 | 20381 | 40.16 s | true · true · recalled | Passed | |
2 · 5 | 20380 | 42.65 s | true · true · recalled | Passed | |
2 · 6 | 20381 | 59.03 s | true · true · recalled | Passed | |
2 · 7 | 20384 | 49.54 s | true · true · recalled | Passed | |
2 · 8 | 20381 | 46.62 s | true · true · recalled | Passed | |
2 · 9 | 20383 | 35.65 s | true · true · recalled | Passed | |
2 · 10 | 20382 | 47.83 s | true · true · recalled | Passed | |
3 · 1 | 20209 | 32.48 s | true · true · recalled | Passed | |
3 · 2 | 20379 | 22.02 s | true · true · recalled | Passed | |
3 · 3 | 20380 | 21.98 s | true · true · recalled | Passed | |
3 · 4 | 20378 | 42.31 s | true · true · recalled | Passed | |
3 · 5 | 20383 | 25.09 s | true · true · recalled | Passed | |
3 · 6 | 20380 | 33.26 s | true · true · recalled | Passed | |
3 · 7 | 20379 | 29.61 s | true · true · recalled | Passed | |
3 · 8 | 20380 | 37.93 s | true · true · recalled | Passed | |
3 · 9 | 20380 | 29.05 s | true · true · recalled | Passed | |
3 · 10 | 20382 | 37.13 s | true · true · recalled | Passed | |
4 · 1 | 20212 | 58.22 s | true · true · recalled | Passed | |
4 · 2 | 20382 | 42.11 s | true · true · recalled | Passed | |
4 · 3 | 20385 | 43.52 s | true · true · recalled | Passed | |
4 · 4 | 20379 | 38.07 s | true · true · recalled | Passed | |
4 · 5 | 20376 | 49.75 s | true · true · recalled | Passed | |
4 · 6 | 20381 | 27.44 s | true · true · recalled | Passed | |
4 · 7 | 20381 | 35.47 s | true · true · recalled | Passed | |
4 · 8 | 20385 | 22.03 s | true · true · recalled | Passed | |
4 · 9 | 20383 | 21.92 s | true · true · recalled | Passed | |
4 · 10 | 20384 | 21.96 s | true · true · recalled | Passed | |
5 · 1 | 20214 | 42.03 s | true · true · recalled | Passed | |
5 · 2 | 20380 | 24.17 s | true · true · recalled | Passed | |
5 · 3 | 20382 | 25.85 s | true · true · recalled | Passed | |
5 · 4 | 20382 | 29.19 s | true · true · recalled | Passed | |
5 · 5 | 20379 | 22.48 s | true · true · recalled | Passed | |
5 · 6 | 20382 | 21.99 s | true · true · recalled | Passed | |
5 · 7 | 20381 | 21.96 s | true · true · recalled | Passed | |
5 · 8 | 20382 | 21.93 s | true · true · recalled | Passed | |
5 · 9 | 20380 | 21.92 s | true · true · recalled | Passed | |
5 · 10 | 20386 | 22.07 s | true · true · recalled | Passed |
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. Bounded literal user-source excerpts, with verified source hashes; details outside them remain unavailable. 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
- Bounded tuning pilot; incomplete window growth cannot pass the full workload grade.
What ran
Exact identity
- Original publisher model
- Qwen3.8-27B
- Publisher revision
1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0- Variant · format · precision
- INT4 variant · safetensors · INT4
- 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 | 61.35 s | per person | max | not recorded | roll_ttft_max_seconds | |
requests ok | 50.00 count | machine total | single | not recorded | requests_ok | |
roll ttft p50 seconds | 34.63 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 | 0.00 count | machine total | single | not recorded | rolled_off_wrong | |
normal ttft max seconds | 58.22 s | per person | max | not recorded | normal_ttft_max_seconds | |
rolled off correct | 45.00 count | machine total | single | not recorded | rolled_off_correct | |
normal ttft p50 seconds | 35.95 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 6abf3f90e1c24c8a8f6c77a9, measured 2026-10-07T20:36:22.746541+00:00. https://skilak.ai/keep/resources/evidence/r/6abf3f90e1c24c8a8f6c77a9
