Model benchmarks

qwen3.8-flash-next-gyro-gyro-s local LLM performance

As of November 2026, qwen3.8-flash-next-gyro-gyro-s runs at up to 68.8 tok/s for local inference (best of 2 community benchmark runs across 2 GPUs).

Strata
ShareRedditX

Model size

Unknown

Peak speed

68.8 tok/s

Average speed

54.9 tok/s

Avg PP

347.6 tok/s

Min memory

36.6 GB

Max context

65,536 tokens

Avg output / run

19,647 tokens

Avg runtime / run

6m 51s

Avg quality

77.1

Benchmark runs

2

GPUs tested

2

Quality by task

Average LLM-judged quality (0–100) with the run-to-run spread shown as a P5–P95 band, overall and for each benchmark task, across all 2 runs. The low and high columns show how much the judge’s score varies between runs, and need at least two runs to display.

TaskP5 (low)AvgP95 (high)
Overall72.477.181.7
Agent Workflow82.186.190.2
Code Generation35.558.781.9
Role Play & Narrative80.986.291.5
Research & Analysis72.577.281.9

Performance by hardware and tool

Every hardware/tool/quantization combination qwen3.8-flash-next-gyro-gyro-s has been benchmarked on, ranked by peak token generation speed. Last updated November 2026.

HardwareToolQuantBest tok/sAvg tok/sMemoryContextQualityRuns
AMD Radeon RX 7900 XTStrata—68.8 tok/s68.8 tok/s36.6 GB65,536 tokens82.21
NVIDIA GeForce RTX 3060Strata—41.1 tok/s41.1 tok/sn/a8,192 tokens71.91

Benchmark runs

All 2 qwen3.8-flash-next-gyro-gyro-s runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.

Frequently asked questions

Is qwen3.8-flash-next-gyro-gyro-s good for coding?
In our benchmarks, qwen3.8-flash-next-gyro-gyro-s scores 58.7/100 for coding. It runs at about 54.9 tok/s, so if you want more speed, turbo is faster (~124.1 tok/s) and still scores well for coding (85.3/100).
Is qwen3.8-flash-next-gyro-gyro-s good for agentic (tool-using) tasks?
In our benchmarks, qwen3.8-flash-next-gyro-gyro-s scores 86.1/100 for agentic workflows. It runs at about 54.9 tok/s, so if you want more speed, granite-4.2-8b-Q4_K_L is faster (~71.6 tok/s) and still scores well for agentic workflows (92.9/100).
How fast is qwen3.8-flash-next-gyro-gyro-s for local inference?
Across 2 community benchmark runs, qwen3.8-flash-next-gyro-gyro-s reaches up to 68.8 tok/s and averages 54.9 tok/s, with the fastest results on AMD Radeon RX 7900 XT.
How much memory does qwen3.8-flash-next-gyro-gyro-s need?
The leanest observed configuration used about 36.6 GB of memory.
Which tools have been used to run qwen3.8-flash-next-gyro-gyro-s?
Benchmarks were submitted using Strata. Results are community-contributed and updated as new runs arrive.