Compare / gpt-oss-20b vs LFM2.5 24B-A2B
gpt-oss-20b vs LFM2.5 24B-A2B
Updated November 2026
Verdict
gpt-oss-20b scores higher on the Small LM Index: 64.0 vs 61.4, although their likely ranges overlap, so the lead is not certain. On the 4 benchmarks both labs report, gpt-oss-20b wins 2 and LFM2.5 24B-A2B wins 2. Both need about 15.7 GB of memory at 4-bit quantization. gpt-oss-20b is cheaper through an API: $0.020 per million tokens against $0.21.
Pick gpt-oss-20b if…
- you want the strongest model overall
- you mostly write code
- API cost matters most
- 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 24B-A2B if…
- you need precise instruction following
Specifications
| gpt-oss-20b | LFM2.5 24B-A2B | |
|---|---|---|
| Small LM Index | 64.0 (likely 60–67) | 61.4 (likely 57–65) |
| Lab | OpenAI · United States | Liquid AI · United States |
| Released | 2025-08-05 | 2026-10-13 |
| Parameters | 20.9B (3.6B active, MoE) | 24B (2B active, MoE) |
| Memory at 4-bit | ~13.8 GB | ~15.7 GB |
| Context window | 131K tokens | 128K tokens |
| Input | text | text |
| Reasoning | Always thinks | No thinking |
| License | Apache-2.0 | LFM Open License |
| API price / 1M tokens | $0.020 ($0.011 in, $0.045 out) | $0.21 ($0.050 in, $0.70 out) |
| Hugging Face downloads (30 days) | 6M | 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 | gpt-oss-20b | LFM2.5 24B-A2B |
|---|---|---|
| MMLU-Pro Knowledge | 75.0* | ≈ 79 |
| GPQA Diamond Knowledge | 71.5 | 74.4 |
| Humanity's Last Exam Reasoning | 10.9 | 7.4 |
| AIME Math | 91.7 | ≈ 85 |
| LiveCodeBench Coding | 61.0* | ≈ 71 |
| SWE-bench Verified Coding | 60.7 | 51.0 |
| IFBench Instruction following | 65.0* | 74.7 |
| 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, gpt-oss-20b or LFM2.5 24B-A2B?
gpt-oss-20b scores higher on the Small LM Index: 64.0 vs 61.4, although their likely ranges overlap, so the lead is not certain. On the 4 benchmarks both labs report, gpt-oss-20b wins 2 and LFM2.5 24B-A2B wins 2.
Which is cheaper to run?
gpt-oss-20b is cheaper through an API: $0.020 per million tokens against $0.21. Both need about 15.7 GB of memory at 4-bit quantization.
Which has the longer context window?
gpt-oss-20b has the longer context window: 131K tokens against 128K.
Can I use gpt-oss-20b and LFM2.5 24B-A2B commercially?
gpt-oss-20b 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.