Model benchmarks
smollm3-3b@q6_k local LLM performance
As of October 2026, smollm3-3b@q6_k runs at up to 33.5 tok/s for local inference (best of 3 community benchmark runs across 1 GPU).
Model size
3B
Peak speed
33.5 tok/s
Average speed
33.0 tok/s
Avg PP
1531.4 tok/s
Min memory
3.6 GB
Max context
32,768 tokens
Avg output / run
14,771 tokens
Avg runtime / run
8m
Avg quality
46.8
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 | 38.8 | 46.8 | 52.6 |
| Agent Workflow | 44.0 | 73.2 | 91.1 |
| Code Generation | 13.6 | 19.2 | 28.4 |
| Role Play & Narrative | 54.2 | 56.2 | 57.5 |
| Research & Analysis | 25.8 | 38.5 | 48.9 |
Performance by hardware and tool
Every hardware/tool/quantization combination smollm3-3b@q6_k 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 2050 | LM Studio | Q6_K | 33.5 tok/s | 33.0 tok/s | 3.6 GB | 32,768 tokens | 46.8 | 3 |
Benchmark runs
All 3 smollm3-3b@q6_k runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is smollm3-3b@q6_k good for coding?
- In our benchmarks, smollm3-3b@q6_k scores 19.2/100 for coding. It runs at about 33.0 tok/s, so if you want more speed, qwen3.8-flash-next-iq3_xxs is faster (~44.6 tok/s) and still scores well for coding (83.0/100).
- Is smollm3-3b@q6_k good for agentic (tool-using) tasks?
- In our benchmarks, smollm3-3b@q6_k scores 73.2/100 for agentic workflows. It runs at about 33.0 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 smollm3-3b@q6_k for local inference?
- Across 3 community benchmark runs, smollm3-3b@q6_k reaches up to 33.5 tok/s and averages 33.0 tok/s, with the fastest results on NVIDIA GeForce RTX 2050.
- How much memory does smollm3-3b@q6_k need?
- The leanest observed configuration used about 3.6 GB of memory (quantizations tested: Q6_K).
- Which tools have been used to run smollm3-3b@q6_k?
- Benchmarks were submitted using LM Studio. Results are community-contributed and updated as new runs arrive.