Model benchmarks
nail-qwen3.6-35b-a3b local LLM performance
As of September 2026, nail-qwen3.6-35b-a3b runs at up to 161.3 tok/s for local inference (best of 4 community benchmark runs across 2 GPUs).
Model size
35B
Peak speed
161.3 tok/s
Average speed
78.5 tok/s
Avg PP
1542.9 tok/s
Min memory
17.0 GB
Max context
131.072 tokens
Avg output / run
22.365 tokens
Avg runtime / run
6m 5s
Avg quality
77.9
Benchmark runs
4
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 4 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 | 76.7 | 77.9 | 79.0 |
| Agent Workflow | 75.2 | 78.8 | 82.2 |
| Code Generation | 72.0 | 72.9 | 74.0 |
| Role Play & Narrative | 82.0 | 85.8 | 90.9 |
| Research & Analysis | 70.6 | 74.1 | 75.9 |
Performance by hardware and tool
Every hardware/tool/quantization combination nail-qwen3.6-35b-a3b 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 |
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 9060 XT | LM Studio | — | 161.3 tok/s | 161.3 tok/s | 17.8 GB | 131.072 tokens | 76.6 | 1 |
| AMD Radeon RX 9050 / 9060 XT | LM Studio | — | 52.1 tok/s | 51.0 tok/s | 17.0 GB | 65.536 tokens | 78.3 | 3 |
Benchmark runs
All 4 nail-qwen3.6-35b-a3b runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is nail-qwen3.6-35b-a3b good for coding?
- In our benchmarks, nail-qwen3.6-35b-a3b scores 72.9/100 for coding. It runs at about 78.5 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 nail-qwen3.6-35b-a3b good for agentic (tool-using) tasks?
- In our benchmarks, nail-qwen3.6-35b-a3b scores 78.8/100 for agentic workflows. It runs at about 78.5 tok/s, so if you want more speed, Tiel-Coder-35B-A3B-MTP-UD-IQ3_XXS is faster (~102.1 tok/s) and still scores well for agentic workflows (88.7/100).
- How fast is nail-qwen3.6-35b-a3b for local inference?
- Across 4 community benchmark runs, nail-qwen3.6-35b-a3b reaches up to 161.3 tok/s and averages 78.5 tok/s, with the fastest results on AMD Radeon RX 9060 XT.
- How much memory does nail-qwen3.6-35b-a3b need?
- The leanest observed configuration used about 17.0 GB of memory.
- Which tools have been used to run nail-qwen3.6-35b-a3b?
- Benchmarks were submitted using LM Studio. Results are community-contributed and updated as new runs arrive.