Compare / LFM2.5 350M vs Gemma 3 270M
LFM2.5 350M vs Gemma 3 270M
Updated November 2026
Verdict
LFM2.5 350M scores higher on the Small LM Index: 16.6 vs 12.8, although their likely ranges overlap, so the lead is not certain. On the 4 benchmarks both labs report, LFM2.5 350M wins 4 and Gemma 3 270M wins 0. Both need about 0.8 GB of memory at 4-bit quantization. Neither model has a hosted API on OpenRouter, so you would run both yourself.
Pick LFM2.5 350M if…
- you want the strongest model overall
- you need precise instruction following
- you build agents that call tools
Pick Gemma 3 270M if…
No clear advantage in the data we have.
Specifications
| LFM2.5 350M | Gemma 3 270M | |
|---|---|---|
| Small LM Index | 16.6 (likely 13–22) | 12.8 (likely 12–14) |
| Lab | Liquid AI · United States | Google · United States |
| Released | 2026-03-31 | 2025-08-14 |
| Parameters | 350M (dense) | 270M (dense) |
| Memory at 4-bit | ~0.8 GB | ~0.7 GB |
| Context window | 32K tokens | 32K tokens |
| Input | text | text |
| Reasoning | No thinking | No thinking |
| License | LFM Open License | Gemma |
| API price / 1M tokens | Self-host only | Self-host only |
| Hugging Face downloads (30 days) | 136.1K | 60.4K |
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 | LFM2.5 350M | Gemma 3 270M |
|---|---|---|
| MMLU-Pro Knowledge | 20.0 | 5.5✓ |
| GPQA Diamond Knowledge | 30.6 | 22.4✓ |
| Humanity's Last Exam Reasoning | ≈ 3 | 3.6✓ |
| AIME Math | ≈ 4 | 2.3✓ |
| LiveCodeBench Coding | ≈ 7 | 0.3✓ |
| τ²-Bench Agentic | 18.4 | 9.1✓ |
| IFBench Instruction following | 40.7 | 12.1✓ |
| IFEval Instruction following | 77.0 | — |
| BFCL Agentic | 21.9 | — |
| AA-LCR Long context | — | 0.0✓ |
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, LFM2.5 350M or Gemma 3 270M?
LFM2.5 350M scores higher on the Small LM Index: 16.6 vs 12.8, although their likely ranges overlap, so the lead is not certain. On the 4 benchmarks both labs report, LFM2.5 350M wins 4 and Gemma 3 270M wins 0.
Which is cheaper to run?
Neither model has a hosted API on OpenRouter, so you would run both yourself. Both need about 0.8 GB of memory at 4-bit quantization.
Which has the longer context window?
Both have a 32K-token context window.
Can I use LFM2.5 350M and Gemma 3 270M commercially?
LFM2.5 350M is released under LFM Open License and Gemma 3 270M under Gemma. Apache 2.0 and MIT allow commercial use; other licenses can add conditions, so read the license before shipping a product.