Compare / Qwen3.5 9B vs LFM2.5 8B-A1B
Qwen3.5 9B vs LFM2.5 8B-A1B
Updated November 2026
Verdict
Qwen3.5 9B scores higher on the Small LM Index: 64.2 vs 40.1. On the 5 benchmarks both labs report, Qwen3.5 9B wins 5 and LFM2.5 8B-A1B wins 0. Both need about 6.4 GB of memory at 4-bit quantization. Only Qwen3.5 9B has a hosted API on OpenRouter ($0.036 per million tokens); the other has to be self-hosted.
Pick Qwen3.5 9B if…
- you want the strongest model overall
- you need precise instruction following
- you build agents that call tools
- you want a hosted API rather than running it yourself
- 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 8B-A1B if…
No clear advantage in the data we have.
Specifications
| Qwen3.5 9B | LFM2.5 8B-A1B | |
|---|---|---|
| Small LM Index | 64.2 (likely 59–69) | 40.1 (likely 37–43) |
| Lab | Alibaba (Qwen) · China | Liquid AI · United States |
| Released | 2026-03-02 | 2026-05-28 |
| Parameters | 9B (dense) | 8.5B (1B active, MoE) |
| Memory at 4-bit | ~6.4 GB | ~5.9 GB |
| Context window | 262K tokens | 128K tokens |
| Input | text, image, video | text |
| Reasoning | Hybrid (thinking on/off) | No thinking |
| License | Apache-2.0 | LFM Open License |
| API price / 1M tokens | $0.036 ($0.032 in, $0.047 out) | Self-host only |
| Hugging Face downloads (30 days) | 6.3M | 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 | Qwen3.5 9B | LFM2.5 8B-A1B |
|---|---|---|
| MMLU-Pro Knowledge | 82.5 | ≈ 63 |
| GPQA Diamond Knowledge | 80.6✓ | 51.3✓ |
| Humanity's Last Exam Reasoning | 14.9✓ | 6.9✓ |
| AIME Math | ≈ 89 | 42.5 |
| LiveCodeBench Coding | 65.6 | ≈ 35 |
| τ²-Bench Agentic | 86.8✓ | 16.1✓ |
| IFBench Instruction following | 66.7✓ | 55.6✓ |
| MMMU-Pro Vision | 69.2✓ | — |
| IFEval Instruction following | — | 91.8 |
| BFCL Agentic | — | 48.5 |
| MATH-500 Math | — | 88.8 |
| AA-LCR Long context | 70.0✓ | 0.0✓ |
| SciCode Coding | 29.5✓ | — |
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, Qwen3.5 9B or LFM2.5 8B-A1B?
Qwen3.5 9B scores higher on the Small LM Index: 64.2 vs 40.1. On the 5 benchmarks both labs report, Qwen3.5 9B wins 5 and LFM2.5 8B-A1B wins 0.
Which is cheaper to run?
Only Qwen3.5 9B has a hosted API on OpenRouter ($0.036 per million tokens); the other has to be self-hosted. Both need about 6.4 GB of memory at 4-bit quantization.
Which has the longer context window?
Qwen3.5 9B has the longer context window: 262K tokens against 128K.
Can I use Qwen3.5 9B and LFM2.5 8B-A1B commercially?
Qwen3.5 9B 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.