Model benchmarks
Qwen3.6-35B-A3B-oQ8e-mtp local LLM performance
As of September 2026, Qwen3.6-35B-A3B-oQ8e-mtp runs at up to 98.6 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
35B
Peak speed
98.6 tok/s
Average speed
95.9 tok/s
Avg prefill
1349.2 tok/s
Min memory
36.7 GB
Max context
262,144 tokens
Avg output / run
14,716 tokens
Avg runtime / run
2m 59s
Avg quality
80.8
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 | 80.2 | 80.8 | 81.3 |
| Agent Workflow | 81.3 | 81.4 | 81.5 |
| Code Generation | 75.7 | 76.7 | 77.7 |
| Role Play & Narrative | 86.0 | 87.2 | 88.3 |
| Research & Analysis | 77.8 | 77.8 | 77.8 |
Performance by hardware and tool
Every hardware/tool/quantization combination Qwen3.6-35B-A3B-oQ8e-mtp 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 M5 Max | oMLX | — | 98.6 tok/s | 95.9 tok/s | 36.7 GB | 262,144 tokens | 80.8 | 2 |
Frequently asked questions
- Is Qwen3.6-35B-A3B-oQ8e-mtp good for coding?
- In our benchmarks, Qwen3.6-35B-A3B-oQ8e-mtp scores 76.7/100 for coding. It runs at about 95.9 tok/s, so if you want more speed, Qwen3.6-35B-A3B-Q4_K_XL is faster (~156.9 tok/s) and still scores well for coding (82.2/100).
- Is Qwen3.6-35B-A3B-oQ8e-mtp good for agentic (tool-using) tasks?
- In our benchmarks, Qwen3.6-35B-A3B-oQ8e-mtp scores 81.4/100 for agentic workflows. It runs at about 95.9 tok/s, so if you want more speed, Tiel-Coder-35B-A3B-Q4_K_S is faster (~165.0 tok/s) and still scores well for agentic workflows (89.3/100).
- How fast is Qwen3.6-35B-A3B-oQ8e-mtp for local inference?
- Across 2 community benchmark runs, Qwen3.6-35B-A3B-oQ8e-mtp reaches up to 98.6 tok/s and averages 95.9 tok/s, with the fastest results on Apple M5 Max.
- How much memory does Qwen3.6-35B-A3B-oQ8e-mtp need?
- The leanest observed configuration used about 36.7 GB of memory.
- Which tools have been used to run Qwen3.6-35B-A3B-oQ8e-mtp?
- Benchmarks were submitted using oMLX. Results are community-contributed and updated as new runs arrive.