Model benchmarks
mudler/Qwen3.6-35B-A3B-APEX-MTP-GGUF local LLM performance
As of September 2026, mudler/Qwen3.6-35B-A3B-APEX-MTP-GGUF runs at up to 31.8 tok/s for local inference (best of 3 community benchmark runs across 1 GPU).
Model size
35B
Peak speed
31.8 tok/s
Average speed
31.7 tok/s
Min memory
17.1 GB
Max context
65,536 tokens
Avg output / run
7,017 tokens
Avg runtime / run
3m 39s
Avg quality
66.4
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 | 64.0 | 66.4 | 69.0 |
| Agent Workflow | 51.8 | 60.2 | 66.4 |
| Code Generation | 62.4 | 66.4 | 69.5 |
| Role Play & Narrative | 65.1 | 65.9 | 66.4 |
| Research & Analysis | 69.0 | 73.1 | 76.2 |
Performance by hardware and tool
Every hardware/tool/quantization combination mudler/Qwen3.6-35B-A3B-APEX-MTP-GGUF 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 |
|---|---|---|---|---|---|---|---|---|
| Advanced Micro Devices, Inc. [AMD/ATI] HawkPoint1 | llama.cpp | — | 31.8 tok/s | 31.7 tok/s | 17.1 GB | 65,536 tokens | 66.4 | 3 |
Frequently asked questions
- Is mudler/Qwen3.6-35B-A3B-APEX-MTP-GGUF good for coding?
- In our benchmarks, mudler/Qwen3.6-35B-A3B-APEX-MTP-GGUF scores 66.4/100 for coding. It runs at about 31.7 tok/s, so if you want more speed, Gemma4:E2B/QAT-MTP@131K is faster (~303.9 tok/s) and still scores well for coding (66.9/100).
- Is mudler/Qwen3.6-35B-A3B-APEX-MTP-GGUF good for agentic (tool-using) tasks?
- In our benchmarks, mudler/Qwen3.6-35B-A3B-APEX-MTP-GGUF scores 60.2/100 for agentic workflows. It runs at about 31.7 tok/s, so if you want more speed, Gemma4:E2B/QAT-MTP@131K is faster (~303.9 tok/s) and still scores well for agentic workflows (71.7/100).
- How fast is mudler/Qwen3.6-35B-A3B-APEX-MTP-GGUF for local inference?
- Across 3 community benchmark runs, mudler/Qwen3.6-35B-A3B-APEX-MTP-GGUF reaches up to 31.8 tok/s and averages 31.7 tok/s, with the fastest results on Advanced Micro Devices, Inc. [AMD/ATI] HawkPoint1.
- How much memory does mudler/Qwen3.6-35B-A3B-APEX-MTP-GGUF need?
- The leanest observed configuration used about 17.1 GB of memory.
- Which tools have been used to run mudler/Qwen3.6-35B-A3B-APEX-MTP-GGUF?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.