- Writing speed, one personsingle_user_decode_tps · tok/s
- 30.70 tok/sper person · median · sample count not recorded
- Machine totalaggregate_e2e_tps · tok/s
- 883.30 tok/smachine total · single · sample count not recorded
Answer checks
168/168 request checks
Method
Streaming load sweep, concurrency 1 through 32, 256-token ceiling; per-request median decode excludes first token; aggregate includes the request window.
Limits
Rented single-GPU class; model and engine pins were not retained. Historical TTFT p95 withdrawn because raw samples cannot reconstruct corrected percentiles.
What ran
Exact identity
- Original publisher model
- Qwen2.5-7B
- Publisher revision
not recorded- Variant · format · precision
- safetensors
- Weights SHA-256
- not recorded
- Engine · version · feature
- vLLM ·
version not recorded· plain - Hardware
- NVIDIA A10, 24 GB
- 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)
- 8192
- 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 |
|---|---|---|---|---|---|---|
requests ok | 168.00 count | machine total | single | not recorded | requests_ok | |
aggregate concurrency | 32.00 count | machine total | single | not recorded | aggregate_concurrency | |
requests total | 168.00 count | machine total | single | not recorded | requests_total | |
aggregate e2e tps | 883.30 tok/s | machine total | single | not recorded | aggregate_tps | |
single user decode tps | 30.70 tok/s | per person | median | not recorded | single_user_decode_tps |
Data and citation
Skilak Keep evidence, run 5cae73717fc898b2755a8d10, measured 2026-07-16. https://skilak.ai/keep/resources/evidence/r/5cae73717fc898b2755a8d10
