#1988 Community Bench: 2026-10-11-community-rtx-3090-qwen3-8-flash-next-unsloth-ud-iq4-xs

Description

[engine-log-excerpt.txt](https://github.com/user-attachments/files/33311385/engine-log-excerpt.txt)
[README.md](https://github.com/user-attachments/files/33311383/README.md)
[results.json](https://github.com/user-attachments/files/33311384/results.json)

# Community benchmark on NVIDIA GeForce RTX 3090

Measured on 2026-10-11 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: 13th Gen Intel(R) Core(TM) i7-13700K (16 cores, 24 threads); installed RAM: 62.5 GiB; storage: not measured
- OS: Linux 7.2.9-2-cachyos; GPU driver: 615.78.08; CUDA: 13.4
- 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: 40000 tokens; parallel slots: 1
- Settings: the install's own configuration, unchanged; thinking off for the suite; temperature 0

```text
strata --pack unsloth-ud-iq4_xs --native Qwen3.8-Flash-Next-UD-IQ4_XS-00001-of-00003.gguf --expert-profile expert-profile.bin --expert-cache auto --prefill auto --spec 4 --mtp rt --max-context 40000 --kv int8 --resident-budget-gib 39 --spec-min-p 0.5
```

## Method

Run by `python tools/strata_bench.py --long` 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 | 64.4 (50.1-73.0) | 0.45 (0.35-0.50) |
| 4K prompt, fresh | 4019 (4019-4019) | 0 (0-0) | 24 cap | 3 | 1396.5 (1202.7-1407.4) | 66.5 (55.0 - 71.7) | 2.89 (2.87-3.36) |
| 16K prompt, fresh | 16058 (16058-16058) | 0 (0-0) | 24 cap | 3 | 2042.9 (2025.9-2089.8) | 74.7 (66.7 - 76.3) | 7.88 (7.70-7.95) |
| 32K prompt, fresh | 32049 (32049-32049) | 0 (0-0) | 24 cap | 3 | 2123.3 (2111.1-2148.8) | 57.9 (51.9 - 68.9) | 15.12 (14.94-15.21) |
| agent session, follow-up turns | 14485 (14482-14523) at the last turn | see below | up to 64 | 3 | see below | 72.3 (65.4-89.2) | 2.59 (2.15-2.78) |

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 % (68.6-68.8); earlier-turn tokens read again per session: 73 (3-73) (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: 1.55 s (1.52-1.63).

| Turn (run 1) | Prompt tokens | Reused | Read | Prompt tok/s | First token (s) | Decode tok/s |
| ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| 0 | 609 | 0 | 609 | 378.6 | 1.632 | 43.1 |
| 1 | 3713 | 633 | 3080 | 1147.5 | 2.717 | 72.3 |
| 2 | 5246 | 3776 | 1470 | 677.3 | 2.191 | 65.7 |
| 3 | 8690 | 5309 | 3381 | 1227.6 | 2.776 | 65.7 |
| 4 | 11614 | 8683 | 2931 | 1100.3 | 2.687 | 74.4 |
| 5 | 14523 | 11679 | 2844 | 1106.8 | 2.591 | 71.1 |

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: 162 s.