My notes so far on Codestral 25.01 - the new code-focused API-only LLM released today by @mistralai.bsky.social simonwillison.net/2025/Jan/13/...
Codestral 25.01
Brand new code-focused model from Mistral. Unlike [the first Codestral](https://simonwillison.net/2024/May/30/codestral/) this one isn't ([yet](https://twitter.com/sophiamyang/status/18789084748114046...
simonwillison.net
January 13, 2025 at 9:45 PM
My notes so far on Codestral 25.01 - the new code-focused API-only LLM released today by @mistralai.bsky.social simonwillison.net/2025/Jan/13/...
Mistral AI Introduces Codestral Embed, the new state-of-the-art embedding model for code.
Codestral Embed can output embeddings with different dimensions and precisions. Codestral Embed with dimension 256 and int8 precision still performs better than any model from our competitors.
Codestral Embed can output embeddings with different dimensions and precisions. Codestral Embed with dimension 256 and int8 precision still performs better than any model from our competitors.
May 29, 2025 at 11:52 PM
Mistral AI Introduces Codestral Embed, the new state-of-the-art embedding model for code.
Codestral Embed can output embeddings with different dimensions and precisions. Codestral Embed with dimension 256 and int8 precision still performs better than any model from our competitors.
Codestral Embed can output embeddings with different dimensions and precisions. Codestral Embed with dimension 256 and int8 precision still performs better than any model from our competitors.
🚀 We just shipped v0.222!
Edit prediction now supports multiple providers: GitHub Copilot's Next Edit Suggestions, Ollama, Mistralai's Codestral, Sweep, and Mercury Coder.
zed.dev/blog/edit-pr...
Edit prediction now supports multiple providers: GitHub Copilot's Next Edit Suggestions, Ollama, Mistralai's Codestral, Sweep, and Mercury Coder.
zed.dev/blog/edit-pr...
Choose Your Edit Prediction Provider - Zed Blog
From the Zed Blog: Zed now supports multiple edit prediction providers: Zeta, Mercury Coder, Sweep, Ollama, and GitHub Copilot Next-Edit.
zed.dev
February 4, 2026 at 6:26 PM
🚀 We just shipped v0.222!
Edit prediction now supports multiple providers: GitHub Copilot's Next Edit Suggestions, Ollama, Mistralai's Codestral, Sweep, and Mercury Coder.
zed.dev/blog/edit-pr...
Edit prediction now supports multiple providers: GitHub Copilot's Next Edit Suggestions, Ollama, Mistralai's Codestral, Sweep, and Mercury Coder.
zed.dev/blog/edit-pr...
Codestral (MistralAI) can now be used as your provider for edit predictions.
Thanks dvcrn!
```
"features": {
"edit_prediction_provider": "codestral"
}
```
Thanks dvcrn!
```
"features": {
"edit_prediction_provider": "codestral"
}
```
October 22, 2025 at 5:43 PM
Codestral (MistralAI) can now be used as your provider for edit predictions.
Thanks dvcrn!
```
"features": {
"edit_prediction_provider": "codestral"
}
```
Thanks dvcrn!
```
"features": {
"edit_prediction_provider": "codestral"
}
```
My notes so far on Codestral 25.01 - the new code-focused API-only LLM released today by Mistral https://simonwillison.net/2025/Jan/13/codestral-2501/
Codestral 25.01
Brand new code-focused model from Mistral. Unlike [the first Codestral](https://simonwillison.net/2024/May/30/codestral/) this one isn't ([yet](https://twitter.com/sophiamyang/status/1878908474811404664)) available as open weights. The model has a 256k token context - a new record for …
simonwillison.net
January 13, 2025 at 9:45 PM
My notes so far on Codestral 25.01 - the new code-focused API-only LLM released today by Mistral https://simonwillison.net/2025/Jan/13/codestral-2501/
Da wären zum einen die moralisch standfesten Menschen hinter Anthropic (Claude-Modelle) und zum anderen europäische Alternativen wie Mistral AI (LeChat-Modelle wie Mistral, Ministral, Devstral oder Codestral).
March 3, 2026 at 3:43 PM
Da wären zum einen die moralisch standfesten Menschen hinter Anthropic (Claude-Modelle) und zum anderen europäische Alternativen wie Mistral AI (LeChat-Modelle wie Mistral, Ministral, Devstral oder Codestral).
Codestral on LM Studio lowkey slays
December 13, 2024 at 1:02 AM
Codestral on LM Studio lowkey slays
Mistral’s new Codestral code completion model races up third-party charts https://venturebeat.com/ai/mistrals-new-codestral-code-completion-model-races-up-third-party-charts/ #AI #coding
Mistral’s new Codestral code completion model races up third-party charts
The newest version of Mistral's code generation model Codestral aims to be more efficient and generate code twice as fast.
venturebeat.com
January 14, 2025 at 1:44 AM
Mistral’s new Codestral code completion model races up third-party charts https://venturebeat.com/ai/mistrals-new-codestral-code-completion-model-races-up-third-party-charts/ #AI #coding
Mistral released their new model - Codestral 25.01 ( mistral.ai/news/codestr... ).
It is doing well on LMSYS Copilot Arena ( lmarena.ai?leaderboard ), but doing just OK in Aider LLM Leaderboard ( aider.chat/docs/leaderb... ).
It is doing well on LMSYS Copilot Arena ( lmarena.ai?leaderboard ), but doing just OK in Aider LLM Leaderboard ( aider.chat/docs/leaderb... ).
January 14, 2025 at 12:46 AM
Mistral released their new model - Codestral 25.01 ( mistral.ai/news/codestr... ).
It is doing well on LMSYS Copilot Arena ( lmarena.ai?leaderboard ), but doing just OK in Aider LLM Leaderboard ( aider.chat/docs/leaderb... ).
It is doing well on LMSYS Copilot Arena ( lmarena.ai?leaderboard ), but doing just OK in Aider LLM Leaderboard ( aider.chat/docs/leaderb... ).
codestral-latest (released Jan 2025) is really fast, but also really dumb (compared to Claude 3.5 Sonnet)
can't wait for new Anthropic releases I guess.
can't wait for new Anthropic releases I guess.
February 24, 2025 at 12:10 PM
codestral-latest (released Jan 2025) is really fast, but also really dumb (compared to Claude 3.5 Sonnet)
can't wait for new Anthropic releases I guess.
can't wait for new Anthropic releases I guess.
Im Bereich der Programmierung habe ich auch bislang vorwiegend mit codex (openAI) gearbeitet, state of the art is jetzt sowieso Claude Code, ich werde mir aber auch Codestral von Mistral näher ansehen. Das wäre nun auch meine präferierte Alternative.
March 6, 2026 at 9:35 AM
Im Bereich der Programmierung habe ich auch bislang vorwiegend mit codex (openAI) gearbeitet, state of the art is jetzt sowieso Claude Code, ich werde mir aber auch Codestral von Mistral näher ansehen. Das wäre nun auch meine präferierte Alternative.
Mistral AIが初のコーディング用生成AIモデル「Codestral」をリリース、80以上のプログラミング言語でトレーニング済み
https://gigazine.net/news/20240530-codestral-mistral-ai/
https://gigazine.net/news/20240530-codestral-mistral-ai/
Mistral AIが初のコーディング用生成AIモデル「Codestral」をリリース、80以上のプログラミング言語でトレーニング済み
MetaとGoogle DeepMindの元従業員が立ち上げたAI開発企業のMistral AIが、同社初のコーディング用AIモデル「Codestral」を発表しました。Codestralはコーディングタスク専用に設計された生成AIモデルで、ソフトウェア開発者が高度なAIアプリケーションを設計するのに役立ちます。
gigazine.net
May 30, 2024 at 3:28 AM
Mistral AIが初のコーディング用生成AIモデル「Codestral」をリリース、80以上のプログラミング言語でトレーニング済み
https://gigazine.net/news/20240530-codestral-mistral-ai/
https://gigazine.net/news/20240530-codestral-mistral-ai/
The only benchmark I've seen that compares DeepSeek v3 and Codestral so far the is the one where DeepSeek v3 got 48% and Codestral got 11% - are there some that I've missed?
January 13, 2025 at 11:43 PM
The only benchmark I've seen that compares DeepSeek v3 and Codestral so far the is the one where DeepSeek v3 got 48% and Codestral got 11% - are there some that I've missed?
I think more people will use mistral chat and API if they release the underlying model weights behind them (especially as models get big like DeepSeek). Goes for new Codestral too.
APIs are still easier to work with, but open releases drive tinkering, research, and tons of engagement.
APIs are still easier to work with, but open releases drive tinkering, research, and tons of engagement.
January 13, 2025 at 3:21 PM
I think more people will use mistral chat and API if they release the underlying model weights behind them (especially as models get big like DeepSeek). Goes for new Codestral too.
APIs are still easier to work with, but open releases drive tinkering, research, and tons of engagement.
APIs are still easier to work with, but open releases drive tinkering, research, and tons of engagement.
Mistral releases its first generative AI model for code
Mistral releases its first generative AI model for code
Mistral, the French AI startup backed by Microsoft and reportedly valued at $6 billion, has released its first generative AI model for coding, dubbed Codestral. Codestral — which is designed to help developers write and interact with code, like the many…
tcrn.ch
May 29, 2024 at 4:58 PM
Mistral releases its first generative AI model for code
Live HTML still lists four Mistral Output rows: OCR $0.0005/page, Medium 3 $2.00, Small 3.1 $0.30, Codestral 2 $0.90. https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing
September 22, 2026 at 5:05 AM
Live HTML still lists four Mistral Output rows: OCR $0.0005/page, Medium 3 $2.00, Small 3.1 $0.30, Codestral 2 $0.90. https://cloud.google.com/gemini-enterprise-agent-platform/generative-ai/pricing
Announcing Codestral 25.08 and the Complete Mistral Coding Stack for Enterprise | Mistral AI
Announcing Codestral 25.08 and the Complete Mistral Coding Stack for Enterprise | Mistral AI
buff.ly
July 30, 2025 at 3:58 PM
Announcing Codestral 25.08 and the Complete Mistral Coding Stack for Enterprise | Mistral AI
Mistral AI unveils Codestral 25.08, optimizing enterprise AI coding with faster, high-fidelity completions and flexible integration in enterprise environments. Full capabilities highlighted here: https://mistral.ai/news/codestral-25-08
July 30, 2025 at 10:01 PM
Mistral AI unveils Codestral 25.08, optimizing enterprise AI coding with faster, high-fidelity completions and flexible integration in enterprise environments. Full capabilities highlighted here: https://mistral.ai/news/codestral-25-08
Mistral's Codestral 25.08
Key improvements include:
🔸+30% increase in accepted completions
🔸+10% more retained code after suggestion
🔸50% fewer runaway generations, improving confidence in longer edits
🔸Improved performance on academic benchmarks for short and long-context FIM completion
Key improvements include:
🔸+30% increase in accepted completions
🔸+10% more retained code after suggestion
🔸50% fewer runaway generations, improving confidence in longer edits
🔸Improved performance on academic benchmarks for short and long-context FIM completion
July 31, 2025 at 2:59 PM
Mistral's Codestral 25.08
Key improvements include:
🔸+30% increase in accepted completions
🔸+10% more retained code after suggestion
🔸50% fewer runaway generations, improving confidence in longer edits
🔸Improved performance on academic benchmarks for short and long-context FIM completion
Key improvements include:
🔸+30% increase in accepted completions
🔸+10% more retained code after suggestion
🔸50% fewer runaway generations, improving confidence in longer edits
🔸Improved performance on academic benchmarks for short and long-context FIM completion
It is true that Mistral never came close to leading anywhere but in specialized models, but those were quite impactful.
Mixtral was the first MoE model that got popular, Codestral the first developer oriented model that was worth looking at.
But Mistral followed up with mediocre models since then.
Mixtral was the first MoE model that got popular, Codestral the first developer oriented model that was worth looking at.
But Mistral followed up with mediocre models since then.
September 8, 2026 at 12:14 PM
It is true that Mistral never came close to leading anywhere but in specialized models, but those were quite impactful.
Mixtral was the first MoE model that got popular, Codestral the first developer oriented model that was worth looking at.
But Mistral followed up with mediocre models since then.
Mixtral was the first MoE model that got popular, Codestral the first developer oriented model that was worth looking at.
But Mistral followed up with mediocre models since then.
Mine tests siger at Claude er bedst til Drupal-moduler efterfulgt af Gemini, ChatGPT og Copilot. Har ikke prøvet med mistral.ai/news/codestral
Codestral | Mistral AI
Empowering developers and democratising coding with Mistral AI.
mistral.ai
July 29, 2025 at 9:53 AM
Mine tests siger at Claude er bedst til Drupal-moduler efterfulgt af Gemini, ChatGPT og Copilot. Har ikke prøvet med mistral.ai/news/codestral
9️⃣ Want to dive deeper?
Check out our full article on Tinz Twins Hub for detailed steps and examples.
tinztwinshub.com/software-eng...
Check out our full article on Tinz Twins Hub for detailed steps and examples.
tinztwinshub.com/software-eng...
Mistral’s Codestral - Create a local AI Coding Assistant for VSCode
Create a local AI coding assistant for VSCode using Mistral’s Codestral model. Save costs and speed up your coding workflow!
tinztwinshub.com
November 22, 2024 at 9:24 PM
9️⃣ Want to dive deeper?
Check out our full article on Tinz Twins Hub for detailed steps and examples.
tinztwinshub.com/software-eng...
Check out our full article on Tinz Twins Hub for detailed steps and examples.
tinztwinshub.com/software-eng...
Mistral AIがコード生成特化型AI「Codestral Mamba」をオープンソースライセンスでリリース
https://gigazine.net/news/20240717-mistral-ai-codestral-mamba/
https://gigazine.net/news/20240717-mistral-ai-codestral-mamba/
Mistral AIがコード生成特化型AI「Codestral Mamba」をオープンソースライセンスでリリース
MetaとGoogle DeepMindの元従業員が立ち上げたAI開発企業のMistral AIが、コーディング用AIモデル「Codestral Mamba」を発表しました。オープンソースライセンスでのリリースであり、商用利用が可能となっています。
gigazine.net
July 17, 2024 at 2:05 AM
Mistral AIがコード生成特化型AI「Codestral Mamba」をオープンソースライセンスでリリース
https://gigazine.net/news/20240717-mistral-ai-codestral-mamba/
https://gigazine.net/news/20240717-mistral-ai-codestral-mamba/
Your code doesn't need to leave your machine to get smart help.
v0.78.0: auto-detect local models (Ollama/LM Studio), one click to add.
Codestral and DeepSeek Coder V2 — top local coding models. Your code stays yours.
ava-supernova.com
#LocalAI #PrivacyFirst #buildinpublic
v0.78.0: auto-detect local models (Ollama/LM Studio), one click to add.
Codestral and DeepSeek Coder V2 — top local coding models. Your code stays yours.
ava-supernova.com
#LocalAI #PrivacyFirst #buildinpublic
June 25, 2026 at 3:00 PM
Your code doesn't need to leave your machine to get smart help.
v0.78.0: auto-detect local models (Ollama/LM Studio), one click to add.
Codestral and DeepSeek Coder V2 — top local coding models. Your code stays yours.
ava-supernova.com
#LocalAI #PrivacyFirst #buildinpublic
v0.78.0: auto-detect local models (Ollama/LM Studio), one click to add.
Codestral and DeepSeek Coder V2 — top local coding models. Your code stays yours.
ava-supernova.com
#LocalAI #PrivacyFirst #buildinpublic