Model benchmarks
Qwen3.8-35B-A3B-Distill.Q4_K_M local LLM performance
As of September 2026, Qwen3.8-35B-A3B-Distill.Q4_K_M runs at up to 41.2 tok/s for local inference (best of 3 community benchmark runs across 1 GPU).
Model size
35B
Peak speed
41.2 tok/s
Average speed
41.2 tok/s
Avg PP
290.9 tok/s
Min memory
17.1 GB
Max context
65.536 tokens
Avg output / run
9.182 tokens
Avg runtime / run
3m 52s
Avg quality
77.8
Benchmark runs
3
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 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 | 75.1 | 77.8 | 81.3 |
| Agent Workflow | 68.3 | 77.3 | 83.8 |
| Code Generation | 66.0 | 69.8 | 73.4 |
| Role Play & Narrative | 87.0 | 89.1 | 90.5 |
| Research & Analysis | 67.9 | 74.9 | 79.4 |
Performance by hardware and tool
Every hardware/tool/quantization combination Qwen3.8-35B-A3B-Distill.Q4_K_M 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 |
|---|---|---|---|---|---|---|---|---|
| Tesla P100-PCIE-16GB | llama.cpp | Q4_K_M | 41.2 tok/s | 41.2 tok/s | 17.1 GB | 65.536 tokens | 77.8 | 3 |
Benchmark runs
All 3 Qwen3.8-35B-A3B-Distill.Q4_K_M runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is Qwen3.8-35B-A3B-Distill.Q4_K_M good for coding?
- In our benchmarks, Qwen3.8-35B-A3B-Distill.Q4_K_M scores 69.8/100 for coding. It runs at about 41.2 tok/s, so if you want more speed, unsloth/Qwen3.8-27B-GGUF:IQ3_S is faster (~104.3 tok/s) and still scores well for coding (83.0/100).
- Is Qwen3.8-35B-A3B-Distill.Q4_K_M good for agentic (tool-using) tasks?
- In our benchmarks, Qwen3.8-35B-A3B-Distill.Q4_K_M scores 77.3/100 for agentic workflows. It runs at about 41.2 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 Qwen3.8-35B-A3B-Distill.Q4_K_M for local inference?
- Across 3 community benchmark runs, Qwen3.8-35B-A3B-Distill.Q4_K_M reaches up to 41.2 tok/s and averages 41.2 tok/s, with the fastest results on Tesla P100-PCIE-16GB.
- How much memory does Qwen3.8-35B-A3B-Distill.Q4_K_M need?
- The leanest observed configuration used about 17.1 GB of memory (quantizations tested: Q4_K_M).
- Which tools have been used to run Qwen3.8-35B-A3B-Distill.Q4_K_M?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.