Model benchmarks
unsloth-qwen3.8:balanced local LLM performance
As of October 2026, unsloth-qwen3.8:balanced runs at up to 38.7 tok/s for local inference (best of 3 community benchmark runs across 0 GPUs).
Model size
Unknown
Peak speed
38.7 tok/s
Average speed
38.1 tok/s
Avg PP
439.1 tok/s
Min memory
n/a
Max context
65,536 tokens
Avg output / run
71,996 tokens
Avg runtime / run
30m 17s
Avg quality
88.1
Benchmark runs
3
GPUs tested
0
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 3 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 | 85.4 | 88.1 | 90.6 |
| Agent Workflow | 90.5 | 90.9 | 91.3 |
| Code Generation | 75.4 | 82.2 | 87.1 |
| Role Play & Narrative | 89.6 | 91.6 | 94.7 |
| Research & Analysis | 84.9 | 87.8 | 90.3 |
Performance by hardware and tool
Every hardware/tool/quantization combination unsloth-qwen3.8:balanced has been benchmarked on, ranked by peak token generation speed. Last updated October 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| AMD Radeon Graphics | llama.cpp | — | 38.7 tok/s | 38.1 tok/s | n/a | 65,536 tokens | 88.1 | 3 |
Benchmark runs
All 3 unsloth-qwen3.8:balanced runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is unsloth-qwen3.8:balanced good for coding?
- In our benchmarks, unsloth-qwen3.8:balanced scores 82.2/100 for coding. It runs at about 38.1 tok/s, so if you want more speed, local-qwen is faster (~60.2 tok/s) and still scores well for coding (84.8/100).
- Is unsloth-qwen3.8:balanced good for agentic (tool-using) tasks?
- In our benchmarks, unsloth-qwen3.8:balanced scores 90.9/100 for agentic workflows. It runs at about 38.1 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 unsloth-qwen3.8:balanced for local inference?
- Across 3 community benchmark runs, unsloth-qwen3.8:balanced reaches up to 38.7 tok/s and averages 38.1 tok/s, with the fastest results on AMD Radeon Graphics.
- Which tools have been used to run unsloth-qwen3.8:balanced?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.