#2097 Community benchmark: RTX 3090, Qwen3.8 Flash-Next UD-IQ4_XS, Strata 0.1.43

Descrição

# Community benchmark on NVIDIA GeForce RTX 3090

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 GeForce RTX 3090, 24 GiB VRAM
- CPU: Intel(R) Xeon(R) CPU E5-2696 v4 @ 2.20GHz (22 cores, 44 threads); installed RAM: 127.8 GiB; storage: not measured
- OS: Windows 11 (10.0.26300); GPU driver: 610.88; CUDA: not read
- Strata: project 0.1.43, commit 7695859a; engine 0.1.43
- Background workloads and power limits: not recorded

## Model and configuration

- Model and quantization: qwen3.8-flash-next-unsloth-ud-iq4_xs
- Context: 262144 tokens; parallel slots: 1
- Settings: the install's own configuration, unchanged; thinking off for the suite; temperature 0

## Method

Run by `python tools/strata_bench.py --long` against the HTTP API of a Strata server (already running); 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 | 43.0 (33.4-53.2) | 0.56 (0.48-0.65) |
| 4K prompt, fresh | 4019 (4019-4019) | 0 (0-0) | 24 cap | 3 | 812.8 (429.3-816.8) | not measured | 4.99 (4.96-9.41) |
| 16K prompt, fresh | 16058 (16058-16058) | 0 (0-0) | 24 cap | 3 | 1365.8 (1005.4-1367.0) | not measured | 11.83 (11.81-16.02) |
| 32K prompt, fresh | 32049 (32049-32049) | 0 (0-0) | 24 cap | 3 | 1426.3 (1421.2-1426.9) | not measured | 22.55 (22.55-22.64) |
| agent session, follow-up turns | 14491 (14487-14493) at the last turn | see below | up to 64 | 3 | see below | 52.6 (41.8-65.5) | 4.82 (4.42-13.52) |

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: 68.7 % (68.7-68.8); earlier-turn tokens read again per session: 74 (4-75) (earlier turns that the cache did not hold; the 13634 (13634-13635) new tokens of the files are read once); first turn (nothing cached) first token: 2.16 s (2.11-3.12).

| Turn (run 1) | Prompt tokens | Reused | Read | Prompt tok/s | First token (s) | Decode tok/s |
| ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| 0 | 609 | 0 | 609 | 197.7 | 3.119 | 33.9 |
| 1 | 3713 | 633 | 3080 | 228.8 | 13.518 | 51.8 |
| 2 | 5246 | 3776 | 1470 | 166.0 | 8.899 | 41.8 |
| 3 | 8690 | 5309 | 3381 | 277.1 | 12.25 | 51.0 |
| 4 | 11614 | 8753 | 2861 | 246.1 | 11.689 | 59.6 |
| 5 | 14487 | 11641 | 2846 | 296.5 | 9.648 | 52.6 |

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: 295 s.