Model benchmarks
llama-3.2-3b-instruct local LLM performance
As of October 2026, llama-3.2-3b-instruct runs at up to 45.9 tok/s for local inference (best of 4 community benchmark runs across 1 GPU).
Model size
3B
Peak speed
45.9 tok/s
Average speed
45.7 tok/s
Avg PP
1462.8 tok/s
Min memory
3.5 GB
Max context
65,536 tokens
Avg output / run
4,059 tokens
Avg runtime / run
1m 35s
Avg quality
34.3
Benchmark runs
4
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 4 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 | 32.3 | 34.3 | 36.6 |
| Agent Workflow | 22.6 | 29.9 | 44.6 |
| Code Generation | 7.0 | 11.6 | 19.3 |
| Role Play & Narrative | 52.2 | 57.6 | 65.1 |
| Research & Analysis | 29.2 | 38.2 | 46.5 |
Performance by hardware and tool
Every hardware/tool/quantization combination llama-3.2-3b-instruct 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 | Q4_K_M | 45.9 tok/s | 45.7 tok/s | 3.5 GB | 65,536 tokens | 34.3 | 4 |
Benchmark runs
All 4 llama-3.2-3b-instruct runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is llama-3.2-3b-instruct good for coding?
- In our benchmarks, llama-3.2-3b-instruct scores 11.6/100 for coding. It runs at about 45.7 tok/s, so if you want more speed, nvidia/Qwen3.8-27B-NVFP4 is faster (~75.5 tok/s) and still scores well for coding (83.5/100).
- Is llama-3.2-3b-instruct good for agentic (tool-using) tasks?
- In our benchmarks, llama-3.2-3b-instruct scores 29.9/100 for agentic workflows. It runs at about 45.7 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 llama-3.2-3b-instruct for local inference?
- Across 4 community benchmark runs, llama-3.2-3b-instruct reaches up to 45.9 tok/s and averages 45.7 tok/s, with the fastest results on NVIDIA GeForce RTX 2050.
- How much memory does llama-3.2-3b-instruct need?
- The leanest observed configuration used about 3.5 GB of memory (quantizations tested: Q4_K_M).
- Which tools have been used to run llama-3.2-3b-instruct?
- Benchmarks were submitted using LM Studio. Results are community-contributed and updated as new runs arrive.