Compare / LFM3 1.7B vs MiniCPM5 2B
LFM3 1.7B vs MiniCPM5 2B
Updated November 2026
Verdict
LFM3 1.7B scores higher on the Small LM Index: 60.0 vs 57.5, although their likely ranges overlap, so the lead is not certain. On the 9 benchmarks both labs report, LFM3 1.7B wins 5 and MiniCPM5 2B wins 4. Both need about 2.4 GB of memory at 4-bit quantization. Neither model has a hosted API on OpenRouter, so you would run both yourself.
Pick LFM3 1.7B if…
- you want the strongest model overall
- you mostly write code
- you need precise instruction following
- you build agents that call tools
- you work with long documents
Pick MiniCPM5 2B if…
- you need a permissive license (Apache-2.0)
- you want the more widely used model (more quantized builds and community support)
Specifications
| LFM3 1.7B | MiniCPM5 2B | |
|---|---|---|
| Small LM Index | 60.0 (likely 58–62) | 57.5 (likely 56–59) |
| Lab | Liquid AI · United States | OpenBMB (MiniCPM) · China |
| Released | 2026-10-27 | 2026-09-06 |
| Parameters | 1.7B (dense) | 2.5B (dense) |
| Memory at 4-bit | ~1.7 GB | ~2.4 GB |
| Context window | 262K tokens | 131K tokens |
| Input | text | text |
| Reasoning | Hybrid (thinking on/off) | Hybrid (thinking on/off) |
| License | LFM Open License | Apache-2.0 |
| API price / 1M tokens | Self-host only | Self-host only |
| Hugging Face downloads (30 days) | 47.8K | 1.1M |
Benchmarks
Higher is better. Bold marks the winner where both labs report a score. ≈ marks the index model's estimate for a score that wasn't reported, ✓ an independent measurement and * a score measured by another lab.
| Benchmark | LFM3 1.7B | MiniCPM5 2B |
|---|---|---|
| MMLU-Pro Knowledge | ≈ 79 | 70.8 |
| GPQA Diamond Knowledge | 63.4 | 70.2 |
| Humanity's Last Exam Reasoning | 7.2 | 8.9 |
| AIME Math | ≈ 84 | 86.5 |
| LiveCodeBench Coding | 73.5 | 69.1 |
| SWE-bench Verified Coding | 43.4 | 46.4 |
| Terminal-Bench Agentic | 12.9 | 8.6 |
| τ²-Bench Agentic | — | 97.1 |
| IFBench Instruction following | 73.0 | 66.3 |
| MMLU Knowledge | — | 84.7 |
| IFEval Instruction following | 96.7 | 86.7 |
| BFCL Agentic | 71.4 | 66.6 |
| MATH-500 Math | — | 94.6 |
| HMMT Math | — | 63.8 |
| AA-LCR Long context | 52.8 | 59.0 |
| SciCode Coding | 28.5 | — |
Scores are mostly reported by each lab with its own settings, so small gaps are noise. See the methodology. Open in the interactive comparison to add more models.
Questions
Which is better, LFM3 1.7B or MiniCPM5 2B?
LFM3 1.7B scores higher on the Small LM Index: 60.0 vs 57.5, although their likely ranges overlap, so the lead is not certain. On the 9 benchmarks both labs report, LFM3 1.7B wins 5 and MiniCPM5 2B wins 4.
Which is cheaper to run?
Neither model has a hosted API on OpenRouter, so you would run both yourself. Both need about 2.4 GB of memory at 4-bit quantization.
Which has the longer context window?
LFM3 1.7B has the longer context window: 262K tokens against 131K.
Can I use LFM3 1.7B and MiniCPM5 2B commercially?
LFM3 1.7B is released under LFM Open License and MiniCPM5 2B under Apache-2.0. Apache 2.0 and MIT allow commercial use; other licenses can add conditions, so read the license before shipping a product.