Compare / Gemma 4 26B-A4B vs LFM2.5 24B-A2B
Gemma 4 26B-A4B vs LFM2.5 24B-A2B
Updated November 2026
Verdict
Gemma 4 26B-A4B scores higher on the Small LM Index: 67.5 vs 61.4. On the 3 benchmarks both labs report, Gemma 4 26B-A4B wins 2 and LFM2.5 24B-A2B wins 1. Both need about 16.8 GB of memory at 4-bit quantization. Gemma 4 26B-A4B is cheaper through an API: $0.068 per million tokens against $0.21.
Pick Gemma 4 26B-A4B if…
- you want the strongest model overall
- API cost matters most
- you work with long documents
- you need a permissive license (Apache-2.0)
- you need image input
- you want the more widely used model (more quantized builds and community support)
Pick LFM2.5 24B-A2B if…
- you mostly write code
Specifications
| Gemma 4 26B-A4B | LFM2.5 24B-A2B | |
|---|---|---|
| Small LM Index | 67.5 (likely 66–69) | 61.4 (likely 57–65) |
| Lab | Google · United States | Liquid AI · United States |
| Released | 2026-04-02 | 2026-10-13 |
| Parameters | 25.2B (3.8B active, MoE) | 24B (2B active, MoE) |
| Memory at 4-bit | ~16.8 GB | ~15.7 GB |
| Context window | 256K tokens | 128K tokens |
| Input | text, image | text |
| Reasoning | Hybrid (thinking on/off) | No thinking |
| License | Apache-2.0 | LFM Open License |
| API price / 1M tokens | $0.068 ($0.040 in, $0.15 out) | $0.21 ($0.050 in, $0.70 out) |
| Hugging Face downloads (30 days) | 5.9M | 32.8K |
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 | Gemma 4 26B-A4B | LFM2.5 24B-A2B |
|---|---|---|
| MMLU-Pro Knowledge | 82.6 | ≈ 79 |
| GPQA Diamond Knowledge | 82.3 | 74.4 |
| Humanity's Last Exam Reasoning | 8.7 | 7.4 |
| AIME Math | 88.3 | ≈ 85 |
| LiveCodeBench Coding | 77.1 | ≈ 71 |
| SWE-bench Verified Coding | 17.4* | 51.0 |
| Terminal-Bench Agentic | 34.2* | — |
| τ²-Bench Agentic | 68.2 | — |
| IFBench Instruction following | — | 74.7 |
| MMMU-Pro Vision | 73.8 | — |
| IFEval Instruction following | — | 96.8 |
| BFCL Agentic | — | 65.0 |
| MATH-500 Math | — | 96.0 |
| SciCode Coding | — | 32.7 |
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, Gemma 4 26B-A4B or LFM2.5 24B-A2B?
Gemma 4 26B-A4B scores higher on the Small LM Index: 67.5 vs 61.4. On the 3 benchmarks both labs report, Gemma 4 26B-A4B wins 2 and LFM2.5 24B-A2B wins 1.
Which is cheaper to run?
Gemma 4 26B-A4B is cheaper through an API: $0.068 per million tokens against $0.21. Both need about 16.8 GB of memory at 4-bit quantization.
Which has the longer context window?
Gemma 4 26B-A4B has the longer context window: 256K tokens against 128K.
Can I use Gemma 4 26B-A4B and LFM2.5 24B-A2B commercially?
Gemma 4 26B-A4B is released under Apache-2.0 and LFM2.5 24B-A2B 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.