#2180 Community benchmark: RTX Pro 4000 Blackwell, qwen3.8-flash-next IQ3_S

Beschreibung

Measured on 2026-10-12 with `tools/strata_bench.py` (suite 1, 3 runs per test, medians with ranges). Speeds are the engine's own timings; first-token times are measured by the client over HTTP. Speed only: no correctness check was run.

## Hardware and software

- GPU and VRAM: NVIDIA RTX PRO 4000 Blackwell, 24 GiB VRAM
- CPU: Intel(R) Xeon(R) Gold 6138 CPU @ 2.00GHz (60 cores, 60 threads); installed RAM: 220.5 GiB; storage: not measured
- OS: Linux 6.12.0-211.49.1.el10_2.x86_64; GPU driver: 580.105.08; CUDA: 13.0
- Strata: project 0.1.43.1, commit a0116b27; engine 0.1.43.1
- Background workloads and power limits: not recorded

## Model and configuration

- Model and quantization: qwen3.8-flash-next-iq3_s
- Context: 131072 tokens; parallel slots: 1
- Settings: the install's own configuration, unchanged; thinking off for the suite; temperature 0

```text
strata --pack iq3_s --native Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00001-of-00002.gguf --ple-gguf Qwen3.8-Flash-Next-GSQ-RCO-IQ3_S-00002-of-00002.gguf --expert-profile expert-profile.bin --expert-cache auto --prefill auto --spec 4 --mtp rt --max-context 131072 --kv int8 --ple-io ram --kv-resident 32768 --vision --vram-reserve-mib 700 --pcie-frac 0.35 --spec-min-p 0.70 --adapt-every 1 --adapt-swaps 160 --adapt-decay 0.97
```

## Method

Run by `python tools/strata_bench.py` against the HTTP API of a Strata server (started by the benchmark from the install configuration); model loading is not timed. One warm-up chat and a 1K prompt are discarded first. Every prompt starts with a unique tag, so the conversation cache cannot help the fresh-prompt tests.

- short chat: the three questions of `tools/ab_engine.py`, up to 100 new tokens, 3 runs
- 4K / 16K prompts: generated word lists sized by a calibration request, 24 new tokens, 3 fresh prompts each
- agent session: system prompt and four tools, then 5 turns that each add about 1500 words of a project file and continue from the model's own earlier answers (the way agent clients resend a conversation), 64 new tokens per turn, 3 sessions; the cache numbers come from the answer's usage (`cached_tokens`) and timings (`prompt_n`)
- concurrency (when run): N requests sent at the same moment, total tokens over the wall time of the burst, latency over all requests

## Results

| Configuration | Actual prompt tokens | Reused tokens | Generated tokens | Runs | Prompt tok/s median and range | Decode tok/s median and range | TTFT seconds median and range |
| --- | ---: | ---: | ---: | ---: | --- | --- | --- |
| short chat | 46 (43-51) | 0 | 100 (100-100) | 9 | not measured | 62.2 (48.4-64.6) | 0.55 (0.46-0.62) |
| 4K prompt, fresh | 4019 (4019-4019) | 0 (0-0) | 24 cap | 3 | 1231.6 (1210.1-1238.6) | not measured | 3.30 (3.28-3.37) |
| 16K prompt, fresh | 16058 (16058-16058) | 0 (0-0) | 24 cap | 3 | 2088.6 (1868.5-2091.2) | not measured | 7.76 (7.75-8.67) |
| agent session, follow-up turns | 14423 (14400-14437) at the last turn | see below | up to 64 | 3 | see below | 75.7 (43.8-105.2) | 3.40 (3.07-3.67) |

Ranges are the lowest and highest of the runs; `Runs` counts requests that returned a token.

### Agent session (conversation cache)

Cache hit over the follow-up turns: 68.6 % (68.6-68.8); earlier-turn tokens read again per session: 74 (3-74) (earlier turns that the cache did not hold; the 13634 (13633-13634) new tokens of the files are read once); first turn (nothing cached) first token: 2.24 s (1.79-2.49).

| Turn (run 1) | Prompt tokens | Reused | Read | Prompt tok/s | First token (s) | Decode tok/s |
| ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| 0 | 609 | 0 | 609 | 247.8 | 2.492 | 37.4 |
| 1 | 3712 | 631 | 3081 | 901.2 | 3.477 | 43.8 |
| 2 | 5245 | 3775 | 1470 | 473.3 | 3.162 | 53.0 |
| 3 | 8689 | 5309 | 3380 | 939.1 | 3.67 | 49.2 |
| 4 | 11574 | 8714 | 2860 | 844.7 | 3.466 | 90.6 |
| 5 | 14437 | 11591 | 2846 | 856.0 | 3.403 | 44.9 |

Per-request data: `results.json`. Engine log lines from the start: `engine-log-excerpt.txt` (when this run started the server).

## Correctness and limitations

Not measured: this is a speed suite. The conversation uses the model's own replies, so token counts differ a little between runs. Machine details come from the server, `nvidia-smi`/`rocm-smi` and the OS; background load and power limits are not recorded. Total time of the suite: 129 s.