Very long inputs
Choose GPT-5-Codex
GPT-5-Codex reads up to 400k tokens per request, against 128k for Codestral.
Codestral vs GPT-5-Codex: price, context and capabilities
Choose GPT-5-Codex for very long inputs, images as input and the latest release; on the other points the two are level or the data is missing.
In short
- A price comparison is not possible: one of them has no comparable published price.
- GPT-5-Codex reads more per request: 400k tokens against 128k.
- GPT-5-Codex is the newer model (Sep 23, 2025, against August 2025).
- Only GPT-5-Codex offers vision, reasoning, structured output, batch api.
Which should you choose?
Each answer below comes straight from the published prices, specifications and independent benchmark results.
Images as input
Choose GPT-5-Codex
GPT-5-Codex accepts images as input; Codestral does not.
Latest release
Choose GPT-5-Codex
GPT-5-Codex is the newer release (Sep 23, 2025, against August 2025 for Codestral).
Specifications side by side
| Codestral | GPT-5-Codex | |
|---|---|---|
| Provider | Mistral AI | OpenAI |
| Category | Language models | Language models |
| Status | Generally available | Deprecated |
| Headline price | No data | No data |
| Input | No data | No data |
| Output | No data | No data |
| Cached input | No data | No data |
| Context window | 128k | 400k |
| Max output | No data | 128k |
| Released | August 2025 | Sep 23, 2025 |
| Knowledge cutoff | No data | September 2024 |
| Licence | Proprietary | Proprietary |
| Capabilities | Tool calling | Vision, Reasoning, Tool calling, Structured output, Batch API |
| API model id | codestral-2508 | gpt-5-codex |
Benchmarks side by side
Leader in bold.
| Benchmark | Codestral | GPT-5-Codex |
|---|---|---|
| Terminal-Bench Hard | No data | 37.9% |
| GPQA Diamond | No data | 83.7% |
| Humanity's Last Exam | No data | 27.8% |
| IFBench (instruction following) | No data | 74.1% |
| τ²-Bench (agentic tool use) | No data | 86.8% |
| MMMU-Pro (visual reasoning) | No data | 73.8% |
| Long-context reasoning (AA-LCR) | No data | 71.7% |
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 Codestral or GPT-5-Codex?
Choose GPT-5-Codex for very long inputs, images as input and the latest release; on the other points the two are level or the data is missing.
Which is cheaper, Codestral or GPT-5-Codex?
A price comparison is not possible: one of them has no comparable published price.
Which has the larger context window, Codestral or GPT-5-Codex?
GPT-5-Codex reads more per request: 400k tokens against 128k.
What is the difference between Codestral and GPT-5-Codex?
Codestral is made by Mistral AI and GPT-5-Codex by OpenAI. A price comparison is not possible: one of them has no comparable published price. GPT-5-Codex reads more per request: 400k tokens against 128k. GPT-5-Codex is the newer model (Sep 23, 2025, against August 2025). Only GPT-5-Codex offers vision, reasoning, structured output, batch api.
Which is newer, Codestral or GPT-5-Codex?
GPT-5-Codex was released on Sep 23, 2025 and Codestral on August 2025.