Overview
Mistral Large (24.02), known on the API as mistral-large-2402, was Mistral AI's first flagship large language model, released on February 26, 2024. It marked the moment the Paris-based startup moved from open 7B/8x7B models into the closed, GPT-4-class tier: Mistral positioned Mistral Large as the second-ranked model generally available through an API at launch, behind only GPT-4. It shipped with a 32K-token context window and native fluency in English, French, Spanish, German, and Italian.
Mistral Large 24.02 was built for reasoning, multilingual work, and developer integration. It supported native function (tool) calling and a JSON output mode that forces responses into valid JSON, making it practical for agentic and structured-extraction pipelines. The launch coincided with the debut of Le Chat, Mistral's free conversational assistant, and a strategic distribution partnership with Microsoft that put Mistral Large on Azure as its first non-Mistral host; it later reached Amazon Bedrock (model ID mistral.mistral-large-2402-v1:0) as well.
This is a legacy model. Mistral AI deprecated mistral-large-2402 on November 30, 2024 and retired it on June 16, 2025, steering users toward the open-weight Mistral Large 2 (24.07/24.11) and then Mistral Large 3. The 24.02 generation itself was never released with open weights — unlike its successor — so it remains a proprietary, API-only artifact of historical interest rather than a model you can deploy today.
| Released | 2024-02-26 |
|---|---|
| License | Proprietary (commercial, closed-source). API-only access — no published model weights. |
| Weights | API only |
| Parameters | Not publicly disclosed by Mistral AI for the 24.02 version |
| Context | 32K tokens |
| Max output | Not officially published by Mistral for the 24.02 release |
| Architecture | Dense transformer (decoder-only). Mistral AI did not publish architecture details or parameter count for the 24.02 release. |
| Knowledge cutoff | Not officially published by Mistral AI |
| Modalities | text input, text output |
| Status | Retired. Deprecated by Mistral AI on 2024-11-30 and retired (end-of-life) on 2025-06-16; superseded by Mistral Large 2 (24.07/24.11) and later Mistral Large 3. |
Benchmarks
- MMLU (5-shot)81.2%
Scores on a 0–100 scale (25-point gridlines); higher is better. Each benchmark links to its published source.
Pricing
| Input | $8.00 per 1M tokens (launch, Feb 2024) per 1M tokens |
|---|---|
| Output | $24.00 per 1M tokens (launch, Feb 2024) per 1M tokens |
Launch pricing per TechCrunch's Feb 26, 2024 report, ~20% below GPT-4 Turbo at the time. Mistral later cut the mistral-large-2402 rate to roughly $4/$12 per 1M tokens before the model was retired on 2025-06-16.
Strengths
- GPT-4-class reasoning for its era — ranked second among API-available models at launch on Mistral's own benchmarks
- Strong multilingual ability with native fluency in English, French, Spanish, German, and Italian
- Native function calling and a strict JSON output mode for agentic and structured workflows
- Broad commercial availability at launch: Mistral La Plateforme, Microsoft Azure (first partner), and later Amazon Bedrock
- Priced roughly 20% below GPT-4 Turbo at launch, making frontier-tier quality cheaper for European-data-residency use cases
Best for
- Complex multilingual reasoning and text generation across English and major European languages
- Structured data extraction and agent tooling via JSON mode and function calling
- Enterprise chat assistants and RAG over documents within the 32K context window
- Code generation and explanation as a general-purpose assistant
- Historical benchmarking and reproducibility studies of early-2024 frontier LLMs
How to access
| Provider | Model ID |
|---|---|
| Mistral AI (La Plateforme) ↗ | mistral-large-2402 |
| Amazon Bedrock ↗ | mistral.mistral-large-2402-v1:0 |
Mistral Large — every version
The full lineage of the Mistral Large line, newest first. Every version has its own page — click any to compare specs, benchmarks and pricing.
| Version | Released | Context | License |
|---|---|---|---|
| Mistral Large 3current | 2025-12-02 | 256K | Apache-2.0 |
| Mistral Large 2.1 (24.11) | 2024-11-18 | — | Open weights |
| Mistral Large 2 (24.07) | 2024-07-24 | — | Open weights |
| Mistral Large (24.02) | 2024-02-26 | — | Proprietary |
FAQ
Is Mistral Large 24.02 still available?
No. Mistral AI deprecated mistral-large-2402 on November 30, 2024 and retired it on June 16, 2025. New projects should use a current model such as Mistral Large 3; the 24.02 endpoint no longer serves requests.
What is the context window of Mistral Large 24.02?
32K tokens (roughly 20,000+ English words). This was the value confirmed in Mistral's launch announcement, its model card, and on Azure and Amazon Bedrock. Note that the later Mistral Large 2 expanded this to 128K tokens.
How much did Mistral Large 24.02 cost?
At its February 2024 launch it was priced at $8 per million input tokens and $24 per million output tokens — about 20% cheaper than GPT-4 Turbo at the time. Mistral later reduced the mistral-large-2402 rate to around $4/$12 per million tokens before retiring it.
Does Mistral Large 24.02 have open weights?
No. The original 24.02 release was a proprietary, API-only model with no published weights. Open weights only arrived with its successor, Mistral Large 2 (24.07), which Mistral released under a research/non-commercial license.