Very long inputs
Choose Grok 4.20 Reasoning
Grok 4.20 Reasoning reads up to 1M tokens per request, against 256k for Mistral Small 4.
Grok 4.20 Reasoning vs Mistral Small 4: price, context and capabilities
Choose Grok 4.20 Reasoning for very long inputs. Choose Mistral Small 4 for self-hosting and data control.
In short
- A price comparison is not possible: one of them has no comparable published price.
- Grok 4.20 Reasoning reads more per request: 1M tokens against 256k.
- Only Grok 4.20 Reasoning offers prompt caching, batch api.
- Only Mistral Small 4 offers self-hostable.
Which should you choose?
Each answer below comes straight from the published prices, specifications and independent benchmark results.
Self-hosting and data control
Choose Mistral Small 4
Mistral Small 4 has open weights, so you can run it on your own servers; Grok 4.20 Reasoning is only available through its provider.
Specifications side by side
| Grok 4.20 Reasoning | Mistral Small 4 | |
|---|---|---|
| Provider | xAI | Mistral AI |
| Category | Language models | Language models |
| Status | Generally available | Generally available |
| Headline price | $1.56 / 1M tokens | No data |
| Input | $1.25 / 1M tokens | No data |
| Output | $2.50 / 1M tokens | No data |
| Cached input | $0.20 / 1M tokens | No data |
| Context window | 1M | 256k |
| Max output | No data | No data |
| Released | March 2026 | March 2026 |
| Knowledge cutoff | No data | No data |
| Licence | Proprietary | Open weights |
| Capabilities | Vision, Reasoning, Tool calling, Structured output, Prompt caching, Batch API | Vision, Reasoning, Tool calling, Structured output, Self-hostable |
| API model id | grok-4.20-0309-reasoning | mistral-small-2603 |
Benchmarks side by side
Leader in bold.
| Benchmark | Grok 4.20 Reasoning | Mistral Small 4 |
|---|---|---|
| Intelligence Index | No data | 11.3 |
| Terminal-Bench Hard | No data | 17.4% |
| SciCode | No data | 38.8% |
| GPQA Diamond | No data | 76.9% |
| Humanity's Last Exam | No data | 9.9% |
| IFBench (instruction following) | No data | 48.2% |
| τ²-Bench (agentic tool use) | No data | 41.2% |
| MMMU-Pro (visual reasoning) | No data | 56.8% |
| Long-context reasoning (AA-LCR) | No data | 49.7% |
Independent evaluations by Artificial Analysis, run the same way for every model, at the reasoning setting shown. Percentages unless marked as an index.
Frequently asked questions
Should I choose Grok 4.20 Reasoning or Mistral Small 4?
Choose Grok 4.20 Reasoning for very long inputs. Choose Mistral Small 4 for self-hosting and data control.
Which is cheaper, Grok 4.20 Reasoning or Mistral Small 4?
A price comparison is not possible: one of them has no comparable published price.
Which has the larger context window, Grok 4.20 Reasoning or Mistral Small 4?
Grok 4.20 Reasoning reads more per request: 1M tokens against 256k.
What is the difference between Grok 4.20 Reasoning and Mistral Small 4?
Grok 4.20 Reasoning is made by xAI and Mistral Small 4 by Mistral AI. A price comparison is not possible: one of them has no comparable published price. Grok 4.20 Reasoning reads more per request: 1M tokens against 256k. Only Grok 4.20 Reasoning offers prompt caching, batch api. Only Mistral Small 4 offers self-hostable.