Model benchmarks
qwen3.6:27b local LLM performance
As of August 2026, qwen3.6:27b runs at up to 38.3 tok/s for local inference (best of 10 community benchmark runs across 2 GPUs).
LM StudioOllamaQ4_K_M
Model size
27.8B
Peak speed
38.3 tok/s
Average speed
19.7 tok/s
Min memory
13.2 GB
Max context
180,000 tokens
Avg runtime / run
9m 6s
Avg quality
79.8
Benchmark runs
10
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 10 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 | 77.4 | 79.8 | 81.9 |
| Agent Workflow | 79.0 | 86.2 | 89.7 |
| Code Generation | 62.6 | 72.3 | 79.2 |
| Role Play & Narrative | 74.9 | 81.6 | 88.3 |
| Research & Analysis | 76.6 | 79.2 | 84.9 |
Performance by hardware and tool
Every hardware/tool/quantization combination qwen3.6:27b has been benchmarked on, ranked by peak token generation speed. Last updated August 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| AMD Radeon RX 7900 XTX | LM Studio | — | 38.3 tok/s | 21.7 tok/s | 16.3 GB | 180,000 tokens | 78.5 | 3 |
| AMD Radeon RX 7900 XTX | Ollama | Q4_K_M | 34.3 tok/s | 23.4 tok/s | 19.0 GB | 65,536 tokens | 80.5 | 5 |
| CPU only | LM Studio | — | 10.6 tok/s | 10.6 tok/s | 13.2 GB | 32,768 tokens | 81.6 | 1 |
| Apple M5 Max | Ollama | Q4_K_M | 4.8 tok/s | 4.8 tok/s | 25.1 GB | 65,536 tokens | 78.4 | 1 |
Frequently asked questions
- Is qwen3.6:27b good for coding?
- In our benchmarks, qwen3.6:27b scores 72.3/100 for coding. It runs at about 19.7 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 qwen3.6:27b good for agentic (tool-using) tasks?
- In our benchmarks, qwen3.6:27b scores 86.2/100 for agentic workflows. It runs at about 19.7 tok/s, so if you want more speed, unsloth/NVIDIA-Nemotron-3.5-Lightning-30B-A3B-GGUF:UD-Q4_K_XL is faster (~169.2 tok/s) and still scores well for agentic workflows (83.0/100).
- How fast is qwen3.6:27b for local inference?
- Across 10 community benchmark runs, qwen3.6:27b reaches up to 38.3 tok/s and averages 19.7 tok/s, with the fastest results on AMD Radeon RX 7900 XTX.
- How much memory does qwen3.6:27b need?
- The leanest observed configuration used about 13.2 GB of memory (quantizations tested: Q4_K_M).
- Which tools have been used to run qwen3.6:27b?
- Benchmarks were submitted using LM Studio, Ollama. Results are community-contributed and updated as new runs arrive.