#2881 Community Benchmark: 1x RTX 5060 Ti 16GB - Qwen3.8-Flash-Next-UD-Q6_K_XL

Descrição

[results.json](https://github.com/user-attachments/files/33593803/results.json)

# Community benchmark on NVIDIA GeForce RTX 5060 Ti

Measured on 2026-10-17 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 5060 Ti, 16 GiB VRAM
- CPU: AMD EPYC 7443 24-Core Processor (24 cores, 24 threads); installed RAM: 224.0 GiB; storage: not measured
- OS: Linux 7.0.6-2-pve; GPU driver: 610.57.04; CUDA: 13.3 (Driver API) / 13.4 (Toolkit)
- Strata: project 0.1.45, commit 477ac13c; engine 0.1.45
- Background workloads and power limits: not recorded

## Model and configuration

- Model and quantization: qwen3.8-flash-next-unsloth-ud-q6_k_xl
- Context: 262144 tokens; parallel slots: 1
- Settings: the install's own configuration, unchanged; thinking off for the suite; temperature 0

```text
strata --pack unsloth-ud-q6_k_xl --native Qwen3.8-Flash-Next-UD-Q6_K_XL-00001-of-00006.gguf --expert-profile expert-profile.bin --expert-cache auto --prefill auto --spec 4 --mtp rt --max-context 262144 --kv int8 --kv-resident 32768 --vision --vram-reserve-mib 700 --pcie-frac 0.00 --spec-min-p 0.70 --adapt-every 1 --adapt-swaps 80 --adapt-decay 0.97
```

Environment of the config: `STRATA_MMQ_KQ=1`, `STRATA_HC_Q8=1`, `STRATA_ROUTE_TAIL_SKIP=0`

## Method

Run by `python tools/strata_bench.py` 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 | 44.7 (38.1-56.4) | 0.64 (0.58-0.75) |
| 4K prompt, fresh | 4019 (4019-4019) | 0 (0-0) | 24 cap | 3 | 498.1 (495.7-498.4) | not measured | 8.10 (8.10-8.14) |
| 16K prompt, fresh | 16058 (16058-16058) | 0 (0-0) | 24 cap | 3 | 937.0 (931.2-938.9) | not measured | 17.18 (17.14-17.28) |
| agent session, follow-up turns | 14573 (14540-14598) at the last turn | see below | up to 64 | 3 | see below | 44.6 (37.6-60.6) | 8.29 (8.02-8.39) |

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: 3 (2-74) (earlier turns that the cache did not hold; the 13717 (13717-13718) new tokens of the files are read once); first turn (nothing cached) first token: 3.37 s (3.32-3.38).

| Turn (run 1) | Prompt tokens | Reused | Read | Prompt tok/s | First token (s) | Decode tok/s |
| ---: | ---: | ---: | ---: | ---: | ---: | ---: |
| 0 | 609 | 0 | 609 | 181.7 | 3.384 | 41.3 |
| 1 | 3767 | 646 | 3121 | 374.7 | 8.363 | 39.9 |
| 2 | 5300 | 3760 | 1540 | 191.9 | 8.063 | 41.3 |
| 3 | 8733 | 5363 | 3370 | 405.2 | 8.357 | 45.6 |
| 4 | 11651 | 8789 | 2862 | 346.1 | 8.308 | 51.8 |
| 5 | 14598 | 11699 | 2899 | 350.0 | 8.321 | 41.3 |

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: 264 s.