Compare / MiniCPM5 2B vs LFM2.5 2.6B
MiniCPM5 2B vs LFM2.5 2.6B
Updated November 2026
Verdict
MiniCPM5 2B scores higher on the Small LM Index: 57.5 vs 43.2. On the 8 benchmarks both labs report, MiniCPM5 2B wins 7 and LFM2.5 2.6B wins 1. Both need about 2.4 GB of memory at 4-bit quantization. Only LFM2.5 2.6B has a hosted API on OpenRouter ($0.39 per million tokens); the other has to be self-hosted.
Pick MiniCPM5 2B if…
- you want the strongest model overall
- you mostly write code
- you need math and reasoning
- you build agents that call tools
- you work with long documents
- you need a permissive license (Apache-2.0)
- you want the more widely used model (more quantized builds and community support)
Pick LFM2.5 2.6B if…
- you want a hosted API rather than running it yourself
Specifications
| MiniCPM5 2B | LFM2.5 2.6B | |
|---|---|---|
| Small LM Index | 57.5 (likely 56–59) | 43.2 (likely 39–48) |
| Lab | OpenBMB (MiniCPM) · China | Liquid AI · United States |
| Released | 2026-09-06 | 2026-07-28 |
| Parameters | 2.5B (dense) | 2.7B (dense) |
| Memory at 4-bit | ~2.4 GB | ~2.3 GB |
| Context window | 131K tokens | 128K tokens |
| Input | text | text |
| Reasoning | Hybrid (thinking on/off) | No thinking |
| License | Apache-2.0 | LFM Open License |
| API price / 1M tokens | Self-host only | $0.39 ($0.35 in, $0.50 out) |
| Hugging Face downloads (30 days) | 1.1M | 69K |
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 | MiniCPM5 2B | LFM2.5 2.6B |
|---|---|---|
| MMLU-Pro Knowledge | 70.8 | ≈ 68 |
| GPQA Diamond Knowledge | 70.2 | 57.8 |
| Humanity's Last Exam Reasoning | 8.9 | 5.9 |
| AIME Math | 86.5 | 51.9 |
| LiveCodeBench Coding | 69.1 | 59.4 |
| SWE-bench Verified Coding | 46.4 | 18.0 |
| Terminal-Bench Agentic | 8.6 | — |
| τ²-Bench Agentic | 97.1 | — |
| IFBench Instruction following | 66.3 | 59.2 |
| MMLU Knowledge | 84.7 | — |
| IFEval Instruction following | 86.7 | 95.4 |
| BFCL Agentic | 66.6 | 56.9 |
| MATH-500 Math | 94.6 | — |
| HMMT Math | 63.8 | — |
| AA-LCR Long context | 59.0 | — |
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, MiniCPM5 2B or LFM2.5 2.6B?
MiniCPM5 2B scores higher on the Small LM Index: 57.5 vs 43.2. On the 8 benchmarks both labs report, MiniCPM5 2B wins 7 and LFM2.5 2.6B wins 1.
Which is cheaper to run?
Only LFM2.5 2.6B has a hosted API on OpenRouter ($0.39 per million tokens); the other has to be self-hosted. Both need about 2.4 GB of memory at 4-bit quantization.
Which has the longer context window?
MiniCPM5 2B has the longer context window: 131K tokens against 128K.
Can I use MiniCPM5 2B and LFM2.5 2.6B commercially?
MiniCPM5 2B is released under Apache-2.0 and LFM2.5 2.6B under LFM Open License. Apache 2.0 and MIT allow commercial use; other licenses can add conditions, so read the license before shipping a product.