Model benchmarks
local/qwen3.8-flash-next local LLM performance
As of September 2026, local/qwen3.8-flash-next runs at up to 74.7 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
Unknown
Peak speed
74.7 tok/s
Average speed
74.6 tok/s
Avg PP
2213.3 tok/s
Min memory
n/a
Max context
8.192 tokens
Avg output / run
9.824 tokens
Avg runtime / run
2m 1s
Avg quality
66.3
Benchmark runs
2
GPUs tested
1
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.
| Task | P5 (low) | Avg | P95 (high) |
|---|---|---|---|
| Overall | 64.1 | 66.3 | 68.5 |
| Agent Workflow | 71.1 | 75.7 | 80.3 |
| Code Generation | 29.5 | 32.7 | 35.8 |
| Role Play & Narrative | 76.0 | 78.4 | 80.8 |
| Research & Analysis | 77.1 | 78.4 | 79.7 |
Performance by hardware and tool
Every hardware/tool/quantization combination local/qwen3.8-flash-next has been benchmarked on, ranked by peak token generation speed. Last updated September 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| Apple M3 Ultra | mlx-serve | — | 74.7 tok/s | 74.6 tok/s | n/a | 8.192 tokens | 66.3 | 2 |
Benchmark runs
All 2 local/qwen3.8-flash-next runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is local/qwen3.8-flash-next good for coding?
- In our benchmarks, local/qwen3.8-flash-next scores 32.7/100 for coding. It runs at about 74.6 tok/s, so if you want more speed, ornith-ai/Ornith-1.5-35B-A3B-GGUF is faster (~96.3 tok/s) and still scores well for coding (83.4/100).
- Is local/qwen3.8-flash-next good for agentic (tool-using) tasks?
- In our benchmarks, local/qwen3.8-flash-next scores 75.7/100 for agentic workflows. It runs at about 74.6 tok/s, so if you want more speed, Ornith-1.5-35B-A3B-BigBang-oQ8e-mtp is faster (~102.7 tok/s) and still scores well for agentic workflows (91.1/100).
- How fast is local/qwen3.8-flash-next for local inference?
- Across 2 community benchmark runs, local/qwen3.8-flash-next reaches up to 74.7 tok/s and averages 74.6 tok/s, with the fastest results on Apple M3 Ultra.
- How much memory does local/qwen3.8-flash-next need?
- The leanest observed configuration used about n/a of memory.
- Which tools have been used to run local/qwen3.8-flash-next?
- Benchmarks were submitted using mlx-serve. Results are community-contributed and updated as new runs arrive.