Model benchmarks
unsloth/Qwen3.6-35B-A3B-MTP-GGUF local LLM performance
As of October 2026, unsloth/Qwen3.6-35B-A3B-MTP-GGUF runs at up to 117.4 tok/s for local inference (best of 5 community benchmark runs across 3 GPUs).
Model size
35B
Peak speed
117.4 tok/s
Average speed
67.2 tok/s
Avg PP
584.9 tok/s
Min memory
15.7 GB
Max context
65,536 tokens
Avg output / run
16,742 tokens
Avg runtime / run
4m 59s
Avg quality
77.6
Benchmark runs
5
GPUs tested
3
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 5 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 | 73.8 | 77.6 | 81.0 |
| Agent Workflow | 68.3 | 78.3 | 86.8 |
| Code Generation | 72.2 | 74.4 | 77.0 |
| Role Play & Narrative | 74.6 | 81.9 | 87.5 |
| Research & Analysis | 67.4 | 75.8 | 82.9 |
Performance by hardware and tool
Every hardware/tool/quantization combination unsloth/Qwen3.6-35B-A3B-MTP-GGUF 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 |
|---|---|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 3090 | Unsloth Studio | UD-Q4_K_XL | 117.4 tok/s | 117.4 tok/s | 26.0 GB | 65,536 tokens | 80.2 | 1 |
| Tesla P100-PCIE-16GB | unsloth | — | 61.8 tok/s | 61.8 tok/s | 15.7 GB | 65,536 tokens | 75.0 | 1 |
| NVIDIA GeForce RTX 3070 Ti | Unsloth Studio | UD-Q4_K_M | 53.0 tok/s | 52.3 tok/s | n/a | 8,192 tokens | 77.6 | 3 |
Benchmark runs
All 5 unsloth/Qwen3.6-35B-A3B-MTP-GGUF runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is unsloth/Qwen3.6-35B-A3B-MTP-GGUF good for coding?
- In our benchmarks, unsloth/Qwen3.6-35B-A3B-MTP-GGUF scores 74.4/100 for coding. It runs at about 67.2 tok/s, so if you want more speed, qwen3.8-27b-Swift1.5-Uncensored-ATX-AI3 is faster (~85.5 tok/s) and still scores well for coding (84.8/100).
- Is unsloth/Qwen3.6-35B-A3B-MTP-GGUF good for agentic (tool-using) tasks?
- In our benchmarks, unsloth/Qwen3.6-35B-A3B-MTP-GGUF scores 78.3/100 for agentic workflows. It runs at about 67.2 tok/s, so if you want more speed, Nex-N2.5-mini-APEX-Mini is faster (~88.2 tok/s) and still scores well for agentic workflows (91.3/100).
- How fast is unsloth/Qwen3.6-35B-A3B-MTP-GGUF for local inference?
- Across 5 community benchmark runs, unsloth/Qwen3.6-35B-A3B-MTP-GGUF reaches up to 117.4 tok/s and averages 67.2 tok/s, with the fastest results on NVIDIA GeForce RTX 3090.
- How much memory does unsloth/Qwen3.6-35B-A3B-MTP-GGUF need?
- The leanest observed configuration used about 15.7 GB of memory (quantizations tested: UD-Q4_K_M, UD-Q4_K_XL).
- Which tools have been used to run unsloth/Qwen3.6-35B-A3B-MTP-GGUF?
- Benchmarks were submitted using Unsloth Studio, unsloth. Results are community-contributed and updated as new runs arrive.