Model benchmarks
qwen_qwen3.5-4b@q8_0 local LLM performance
As of November 2026, qwen_qwen3.5-4b@q8_0 runs at up to 69.0 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
4B
Peak speed
69.0 tok/s
Average speed
68.4 tok/s
Avg PP
599.1 tok/s
Min memory
6.5 GB
Max context
65,536 tokens
Avg output / run
17,928 tokens
Avg runtime / run
4m 36s
Avg quality
69.9
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 | 67.1 | 69.9 | 72.7 |
| Agent Workflow | 73.0 | 73.9 | 74.9 |
| Code Generation | 45.1 | 55.9 | 66.6 |
| Role Play & Narrative | 74.6 | 78.3 | 82.1 |
| Research & Analysis | 66.1 | 71.3 | 76.5 |
Performance by hardware and tool
Every hardware/tool/quantization combination qwen_qwen3.5-4b@q8_0 has been benchmarked on, ranked by peak token generation speed. Last updated November 2026.
| Hardware | Tool | Quant | Best tok/s | Avg tok/s | Memory | Context | Quality | Runs |
|---|---|---|---|---|---|---|---|---|
| Apple M1 Max | LM Studio | Q8_0 | 69.0 tok/s | 68.4 tok/s | 6.5 GB | 65,536 tokens | 69.9 | 2 |
Benchmark runs
All 2 qwen_qwen3.5-4b@q8_0 runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is qwen_qwen3.5-4b@q8_0 good for coding?
- In our benchmarks, qwen_qwen3.5-4b@q8_0 scores 55.9/100 for coding. It runs at about 68.4 tok/s, so if you want more speed, ornith-ai/Ornith-1.5-35B-A3B-GGUF is faster (~80.3 tok/s) and still scores well for coding (82.1/100).
- Is qwen_qwen3.5-4b@q8_0 good for agentic (tool-using) tasks?
- In our benchmarks, qwen_qwen3.5-4b@q8_0 scores 73.9/100 for agentic workflows. It runs at about 68.4 tok/s, so if you want more speed, swift-iq2_xs is faster (~91.5 tok/s) and still scores well for agentic workflows (91.4/100).
- How fast is qwen_qwen3.5-4b@q8_0 for local inference?
- Across 2 community benchmark runs, qwen_qwen3.5-4b@q8_0 reaches up to 69.0 tok/s and averages 68.4 tok/s, with the fastest results on Apple M1 Max.
- How much memory does qwen_qwen3.5-4b@q8_0 need?
- The leanest observed configuration used about 6.5 GB of memory (quantizations tested: Q8_0).
- Which tools have been used to run qwen_qwen3.5-4b@q8_0?
- Benchmarks were submitted using LM Studio. Results are community-contributed and updated as new runs arrive.