- Code writing speedcode_decode_tps · tok/s
- 45.22 tok/sper person · mean · sample count not recorded
- Prose writing speedchat_decode_tps · tok/s
- 27.70 tok/sper person · mean · sample count not recorded
- Peak memory, whole hostpeak_host_unavailable_gib · GiB
- 92.44 GiBmachine total · max · sample count not recorded
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 |
|---|---|---|---|---|---|---|
2026-10-05 | 2 people | not recorded | not recorded | not recorded | Not graded | |
2026-10-05 | 1 people | not recorded | not recorded | not recorded | Not graded |
Answer checks
6/6 request checks
Method
One warmup then three measured requests per prompt at concurrency 1; completion tokens divided by first-content-to-DONE time; unchanged caches.
Limits
Short repeated code/chat prompts; output hit the token ceiling; task correctness and full context quality unassessed.
What ran
Exact identity
- Original publisher model
- Qwen3.8-27B
- Publisher revision
1d4bf0f2ff6012fd82039f2fa52739d0dd7c60c0- Variant · format · precision
- NVFP4 variant with BF16 head · safetensors · NVFP4
- Weights SHA-256
- not recorded
- Engine · version · feature
- vLLM ·
version not recorded· speculative decoding - Hardware
- NVIDIA GB10, 128 GB unified memory
- Harness / workload
not recorded- Workload suite · version
not recorded- Harness SHA-256
not recorded- Fixture SHA-256
not recorded
Workload shape
- People in this run
- 1
- Configured window (tokens)
- 262144
- KV cache precision
- fp8
- Output token ceiling
- 512
- Feature
- speculative decoding
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 |
|---|---|---|---|---|---|---|
chat decode tps | 27.70 tok/s | per person | mean | not recorded | chat_decode_tps | |
requests ok | 6.00 count | machine total | single | not recorded | requests_ok | |
requests total | 6.00 count | machine total | single | not recorded | requests_total | |
code decode tps | 45.22 tok/s | per person | mean | not recorded | code_decode_tps | |
peak host unavailable gib | 92.44 GiB | machine total | max | not recorded | peak_host_unavailable_gib |
Data and citation
Skilak Keep evidence, run 60cee206868b13f9f5272ee1, measured 2026-10-05. https://skilak.ai/keep/resources/evidence/r/60cee206868b13f9f5272ee1
