Compare / Mellum2.1 12B-A2.5B Thinking vs Mellum2 12B-A2.5B Instruct
Mellum2.1 12B-A2.5B Thinking vs Mellum2 12B-A2.5B Instruct
Updated November 2026
Verdict
Mellum2.1 12B-A2.5B Thinking scores higher on the Small LM Index: 60.8 vs 39.2. On the 6 benchmarks both labs report, Mellum2.1 12B-A2.5B Thinking wins 6 and Mellum2 12B-A2.5B Instruct wins 0. Both need about 8.3 GB of memory at 4-bit quantization. Neither model has a hosted API on OpenRouter, so you would run both yourself.
Pick Mellum2.1 12B-A2.5B Thinking if…
- you want the strongest model overall
- you mostly write code
- you need math and reasoning
- you need precise instruction following
- you build agents that call tools
Pick Mellum2 12B-A2.5B Instruct if…
No clear advantage in the data we have.
Specifications
| Mellum2.1 12B-A2.5B Thinking | Mellum2 12B-A2.5B Instruct | |
|---|---|---|
| Small LM Index | 60.8 (likely 58–63) | 39.2 (likely 36–42) |
| Lab | JetBrains (Mellum) · Europe | JetBrains (Mellum) · Europe |
| Released | 2026-09-20 | 2026-06-01 |
| Parameters | 12.2B (2.5B active, MoE) | 12.2B (2.5B active, MoE) |
| Memory at 4-bit | ~8.3 GB | ~8.3 GB |
| Context window | 131K tokens | 131K tokens |
| Input | text | text |
| Reasoning | Always thinks | No thinking |
| License | Apache-2.0 | Apache-2.0 |
| API price / 1M tokens | Self-host only | Self-host only |
| Hugging Face downloads (30 days) | 10.5K | 11.1K |
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 | Mellum2.1 12B-A2.5B Thinking | Mellum2 12B-A2.5B Instruct |
|---|---|---|
| GPQA Diamond Knowledge | 64.6 | 40.9 |
| AIME Math | 83.3 | 41.7 |
| LiveCodeBench Coding | 82.0 | 37.2 |
| SWE-bench Verified Coding | 47.0 | ≈ 42 |
| Terminal-Bench Agentic | 17.4 | — |
| MMLU Knowledge | 87.8 | 78.1 |
| IFEval Instruction following | 90.6 | 75.8 |
| BFCL Agentic | 62.3 | 44.2 |
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, Mellum2.1 12B-A2.5B Thinking or Mellum2 12B-A2.5B Instruct?
Mellum2.1 12B-A2.5B Thinking scores higher on the Small LM Index: 60.8 vs 39.2. On the 6 benchmarks both labs report, Mellum2.1 12B-A2.5B Thinking wins 6 and Mellum2 12B-A2.5B Instruct wins 0.
Which is cheaper to run?
Neither model has a hosted API on OpenRouter, so you would run both yourself. Both need about 8.3 GB of memory at 4-bit quantization.
Which has the longer context window?
Both have a 131K-token context window.
Can I use Mellum2.1 12B-A2.5B Thinking and Mellum2 12B-A2.5B Instruct commercially?
Both are released under the Apache-2.0 license. Apache 2.0 and MIT allow commercial use; other licenses can add conditions, so read the license before shipping a product.