Model benchmarks
deepseek-r1-0528-qwen3-8b local LLM performance
As of September 2026, deepseek-r1-0528-qwen3-8b runs at up to 42.8 tok/s for local inference (best of 2 community benchmark runs across 1 GPU).
Model size
8B
Peak speed
42.8 tok/s
Average speed
37.3 tok/s
Avg PP
552.9 tok/s
Min memory
4.7 GB
Max context
65.536 tokens
Avg output / run
27.500 tokens
Avg runtime / run
13m 8s
Avg quality
30.6
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 | 28.9 | 30.6 | 32.2 |
| Agent Workflow | 29.4 | 45.1 | 60.9 |
| Code Generation | 3.8 | 6.5 | 9.2 |
| Role Play & Narrative | 45.2 | 57.7 | 70.3 |
| Research & Analysis | 12.3 | 12.9 | 13.5 |
Performance by hardware and tool
Every hardware/tool/quantization combination deepseek-r1-0528-qwen3-8b 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 |
|---|---|---|---|---|---|---|---|---|
| Apple M4 Max | LM Studio | — | 42.8 tok/s | 42.8 tok/s | 4.7 GB | 65.536 tokens | 32.4 | 1 |
| CPU only | LM Studio | — | 31.9 tok/s | 31.9 tok/s | 4.7 GB | 8.192 tokens | 28.8 | 1 |
Benchmark runs
All 2 deepseek-r1-0528-qwen3-8b runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is deepseek/deepseek-r1-0528-qwen3-8b good for coding?
- In our benchmarks, deepseek/deepseek-r1-0528-qwen3-8b scores 6.5/100 for coding. It runs at about 37.3 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 deepseek/deepseek-r1-0528-qwen3-8b good for agentic (tool-using) tasks?
- In our benchmarks, deepseek/deepseek-r1-0528-qwen3-8b scores 45.1/100 for agentic workflows. It runs at about 37.3 tok/s, so if you want more speed, granite-4.2-8b-Q4_K_L is faster (~71.6 tok/s) and still scores well for agentic workflows (92.9/100).
- How fast is deepseek-r1-0528-qwen3-8b for local inference?
- Across 2 community benchmark runs, deepseek-r1-0528-qwen3-8b reaches up to 42.8 tok/s and averages 37.3 tok/s, with the fastest results on Apple M4 Max.
- How much memory does deepseek-r1-0528-qwen3-8b need?
- The leanest observed configuration used about 4.7 GB of memory.
- Which tools have been used to run deepseek-r1-0528-qwen3-8b?
- Benchmarks were submitted using LM Studio. Results are community-contributed and updated as new runs arrive.