Compare / Magistral Small 1.2 vs Tri-21B Think
Magistral Small 1.2 vs Tri-21B Think
Updated November 2026
Verdict
Magistral Small 1.2 scores higher on the Small LM Index: 55.9 vs 43.6. On the 3 benchmarks both labs report, Magistral Small 1.2 wins 3 and Tri-21B Think wins 0. Both need about 16.9 GB of memory at 4-bit quantization. Neither model has a hosted API on OpenRouter, so you would run both yourself.
Pick Magistral Small 1.2 if…
- you want the strongest model overall
- you mostly write code
- you need math and reasoning
- you work with long documents
- you need image input
- you want the more widely used model (more quantized builds and community support)
Pick Tri-21B Think if…
No clear advantage in the data we have.
Specifications
| Magistral Small 1.2 | Tri-21B Think | |
|---|---|---|
| Small LM Index | 55.9 (likely 51–60) | 43.6 (likely 40–48) |
| Lab | Mistral AI · Europe | Trillion Labs · Rest of Asia |
| Released | 2025-09-17 | 2026-02-18 |
| Parameters | 24B (dense) | 20.7B (dense) |
| Memory at 4-bit | ~16.9 GB | ~15.1 GB |
| Context window | 128K tokens | 32K tokens |
| Input | text, image | text |
| Reasoning | Always thinks | Always thinks |
| License | Apache-2.0 | Apache-2.0 |
| API price / 1M tokens | Self-host only | Self-host only |
| Hugging Face downloads (30 days) | 29.7K | 459 |
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 | Magistral Small 1.2 | Tri-21B Think |
|---|---|---|
| MMLU-Pro Knowledge | ≈ 76 | 74.3 |
| GPQA Diamond Knowledge | 70.1 | 62.6 |
| Humanity's Last Exam Reasoning | ≈ 9 | 5.5 |
| AIME Math | 77.3 | 56.7 |
| LiveCodeBench Coding | 70.9 | 53.7 |
| IFBench Instruction following | — | 56.7 |
| IFEval Instruction following | — | 84.7 |
| AA-LCR Long context | — | 11.0 |
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, Magistral Small 1.2 or Tri-21B Think?
Magistral Small 1.2 scores higher on the Small LM Index: 55.9 vs 43.6. On the 3 benchmarks both labs report, Magistral Small 1.2 wins 3 and Tri-21B Think wins 0.
Which is cheaper to run?
Neither model has a hosted API on OpenRouter, so you would run both yourself. Both need about 16.9 GB of memory at 4-bit quantization.
Which has the longer context window?
Magistral Small 1.2 has the longer context window: 128K tokens against 32K.
Can I use Magistral Small 1.2 and Tri-21B Think 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.