Compare / Qwen3 8B vs LFM2.5 8B-A1B
Qwen3 8B vs LFM2.5 8B-A1B
Updated November 2026
Verdict
Qwen3 8B scores higher on the Small LM Index: 53.9 vs 40.1. On the 7 benchmarks both labs report, Qwen3 8B wins 4 and LFM2.5 8B-A1B wins 2. Both need about 6.8 GB of memory at 4-bit quantization. Neither model has a hosted API on OpenRouter, so you would run both yourself.
Pick Qwen3 8B if…
- you want the strongest model overall
- 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 8B-A1B if…
- you need precise instruction following
Specifications
| Qwen3 8B | LFM2.5 8B-A1B | |
|---|---|---|
| Small LM Index | 53.9 (likely 51–57) | 40.1 (likely 37–43) |
| Lab | Alibaba (Qwen) · China | Liquid AI · United States |
| Released | 2025-04-29 | 2026-05-28 |
| Parameters | 8.2B (dense) | 8.5B (1B active, MoE) |
| Memory at 4-bit | ~6.8 GB | ~5.9 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 | Self-host only |
| Hugging Face downloads (30 days) | 7.3M | 31.3K |
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 | Qwen3 8B | LFM2.5 8B-A1B |
|---|---|---|
| MMLU-Pro Knowledge | 74.3✓ | ≈ 63 |
| GPQA Diamond Knowledge | 58.9✓ | 51.3✓ |
| Humanity's Last Exam Reasoning | 3.9✓ | 6.9✓ |
| AIME Math | 74.7✓ | 42.5 |
| LiveCodeBench Coding | 40.6✓ | ≈ 35 |
| τ²-Bench Agentic | 27.8✓ | 16.1✓ |
| IFBench Instruction following | 33.5✓ | 55.6✓ |
| IFEval Instruction following | — | 91.8 |
| BFCL Agentic | — | 48.5 |
| MATH-500 Math | 90.4✓ | 88.8 |
| AA-LCR Long context | 0.0✓ | 0.0✓ |
| SciCode Coding | 27.9✓ | — |
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, Qwen3 8B or LFM2.5 8B-A1B?
Qwen3 8B scores higher on the Small LM Index: 53.9 vs 40.1. On the 7 benchmarks both labs report, Qwen3 8B wins 4 and LFM2.5 8B-A1B wins 2.
Which is cheaper to run?
Neither model has a hosted API on OpenRouter, so you would run both yourself. Both need about 6.8 GB of memory at 4-bit quantization.
Which has the longer context window?
Qwen3 8B has the longer context window: 131K tokens against 128K.
Can I use Qwen3 8B and LFM2.5 8B-A1B commercially?
Qwen3 8B is released under Apache-2.0 and LFM2.5 8B-A1B 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.