Model benchmarks
lfm3-1.7b local LLM performance
As of November 2026, lfm3-1.7b runs at up to 84.3 tok/s for local inference (best of 2 community benchmark runs across 2 GPUs).
Model size
1.7B
Peak speed
84.3 tok/s
Average speed
80.4 tok/s
Avg PP
1687.7 tok/s
Min memory
4.4 GB
Max context
65,536 tokens
Avg output / run
29,897 tokens
Avg runtime / run
6m 25s
Avg quality
55.6
Benchmark runs
2
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 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 | 52.4 | 55.6 | 58.8 |
| Agent Workflow | 76.3 | 80.6 | 84.9 |
| Code Generation | 10.6 | 25.4 | 40.1 |
| Role Play & Narrative | 52.0 | 55.7 | 59.5 |
| Research & Analysis | 50.5 | 60.6 | 70.7 |
Performance by hardware and tool
Every hardware/tool/quantization combination lfm3-1.7b 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 |
|---|---|---|---|---|---|---|---|---|
| NVIDIA GeForce RTX 3060 | llama.cpp | — | 84.3 tok/s | 84.3 tok/s | 4.4 GB | 8,192 tokens | 52.0 | 1 |
| Tesla P100-PCIE-16GB | llama.cpp | — | 76.5 tok/s | 76.5 tok/s | 5.3 GB | 65,536 tokens | 59.1 | 1 |
Benchmark runs
All 2 lfm3-1.7b runs submitted so far — expand one for its hardware, quality breakdown and per-scenario detail.
Frequently asked questions
- Is LFM3-1.7B good for coding?
- In our benchmarks, LFM3-1.7B scores 25.4/100 for coding. It runs at about 80.4 tok/s, so if you want more speed, Tiel-Coder-35B-A3B-MLX-oQ4e is faster (~97.4 tok/s) and still scores well for coding (80.9/100).
- Is LFM3-1.7B good for agentic (tool-using) tasks?
- In our benchmarks, LFM3-1.7B scores 80.6/100 for agentic workflows. It runs at about 80.4 tok/s, so if you want more speed, qwen3.8-flash-next-iq3_s is faster (~98.3 tok/s) and still scores well for agentic workflows (89.9/100).
- How fast is lfm3-1.7b for local inference?
- Across 2 community benchmark runs, lfm3-1.7b reaches up to 84.3 tok/s and averages 80.4 tok/s, with the fastest results on NVIDIA GeForce RTX 3060.
- How much memory does lfm3-1.7b need?
- The leanest observed configuration used about 4.4 GB of memory.
- Which tools have been used to run lfm3-1.7b?
- Benchmarks were submitted using llama.cpp. Results are community-contributed and updated as new runs arrive.