Compare / Muse Glimmer 30B vs Kanana-2 30B-A3B Thinking
Muse Glimmer 30B vs Kanana-2 30B-A3B Thinking
Updated November 2026
Verdict
Muse Glimmer 30B scores higher on the Small LM Index: 72.6 vs 50.7. On the 3 benchmarks both labs report, Muse Glimmer 30B wins 3 and Kanana-2 30B-A3B Thinking wins 0. Both need about 20.2 GB of memory at 4-bit quantization. Only Muse Glimmer 30B has a hosted API on OpenRouter ($0.53 per million tokens); the other has to be self-hosted.
Pick Muse Glimmer 30B if…
- you want the strongest model overall
- you need math and reasoning
- you need precise instruction following
- 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 Kanana-2 30B-A3B Thinking if…
No clear advantage in the data we have.
Specifications
| Muse Glimmer 30B | Kanana-2 30B-A3B Thinking | |
|---|---|---|
| Small LM Index | 72.6 (likely 71–74) | 50.7 (likely 46–55) |
| Lab | Meta · United States | Kakao (Kanana) · Rest of Asia |
| Released | 2026-08-09 | 2026-01-14 |
| Parameters | 29.6B (dense) | 30.7B (3B active, MoE) |
| Memory at 4-bit | ~19.2 GB | ~20.2 GB |
| Context window | 131K tokens | 32K tokens |
| Input | text, image | text |
| Reasoning | Always thinks | Always thinks |
| License | Apache-2.0 | Kanana License |
| API price / 1M tokens | $0.53 ($0.30 in, $1.20 out) | Self-host only |
| Hugging Face downloads (30 days) | 189.7K | 1.3K |
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 | Muse Glimmer 30B | Kanana-2 30B-A3B Thinking |
|---|---|---|
| MMLU-Pro Knowledge | ≈ 86 | 74.2 |
| GPQA Diamond Knowledge | 83.5✓ | 57.8 |
| Humanity's Last Exam Reasoning | 22.0✓ | ≈ 7 |
| AIME Math | 94.7 | 74.0 |
| LiveCodeBench Coding | ≈ 83 | 58.8 |
| SWE-bench Verified Coding | 76.0 | ≈ 32 |
| Terminal-Bench Agentic | 51.7 | — |
| IFBench Instruction following | 77.0 | 47.8 |
| MMMU-Pro Vision | 74.0 | — |
| IFEval Instruction following | — | 82.2 |
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, Muse Glimmer 30B or Kanana-2 30B-A3B Thinking?
Muse Glimmer 30B scores higher on the Small LM Index: 72.6 vs 50.7. On the 3 benchmarks both labs report, Muse Glimmer 30B wins 3 and Kanana-2 30B-A3B Thinking wins 0.
Which is cheaper to run?
Only Muse Glimmer 30B has a hosted API on OpenRouter ($0.53 per million tokens); the other has to be self-hosted. Both need about 20.2 GB of memory at 4-bit quantization.
Which has the longer context window?
Muse Glimmer 30B has the longer context window: 131K tokens against 32K.
Can I use Muse Glimmer 30B and Kanana-2 30B-A3B Thinking commercially?
Muse Glimmer 30B is released under Apache-2.0 and Kanana-2 30B-A3B Thinking under Kanana License. Apache 2.0 and MIT allow commercial use; other licenses can add conditions, so read the license before shipping a product.