Science & Tech · AI · 7 hrs ago
Mistral Says Its Open-Weight AI Model Leads in Coding and Cybersecurity
French AI company Mistral has released its new model, Large 4, which it nicknames Le Chonk.
The company says the model, with one trillion parameters, is its most powerful so far and now leads open-weight models in automated coding and cybersecurity.
Mistral says it matches or surpasses leading Chinese models after training with increasingly difficult coding and cybersecurity problems.
The finalized model is freely available, and its weights can be downloaded and adapted rather than controlled remotely by the company.
That openness may help businesses and governments use AI for cyber defense without depending on a service that could be switched off, including one from another country.
But downloadable models can also be misused, and a security expert says advanced AI from any country needs careful monitoring and controls.
Mistral is also positioning itself as a European alternative to US and Chinese AI companies, though experts say it is difficult to compare models and that US firms still offer the most powerful ones.
Mistral says its finalized Large 4 model, nicknamed Le Chonk, leads open-weight models in automated coding and cybersecurity.
The French AI company released the model’s weights for public access on October 29.
Mistral says the model improved after weeks of training on harder coding and cybersecurity problems.
ZDNet reports that Large 4 scored 41% on Terminal-Bench 4.0, while its overall Coding Agent Index score tied Codex GPT-6.1 Terra.
Security experts warn that open-weight models can be abused as well as used for defense.
- Who
- French AI company Mistral released Large 4, nicknamed Le Chonk.
- What
- Mistral says the open-weight model now leads in automated coding and cybersecurity.
- When
- October 29, 2026. Mistral had released an early version on October 6.
- Where
- The finalized model is freely available to access; a specific release location is Not stated.
- Why
- Mistral says the model can help enterprises and governments defend against cyber threats and reduce dependence on proprietary or foreign AI.
Mistral
Security expert Rafe Pilling
Open-weight access
Mistral
Mistral says freely available models can help businesses and governments defend against cyber threats and cannot be switched off remotely.
Security expert Rafe Pilling
Pilling says models available to download and circulate without restriction can be abused as well as used productively.
Security risks
Mistral
Mistral argues that open models can support cyber defense and user control.
Security expert Rafe Pilling
Pilling says advanced models can go rogue without careful supervision, regardless of their country of origin, and monitoring and controls are needed.
Mistral released an early preview of Large 4, nicknamed Le Chonk.
Mistral released the finalized model’s weights and reported new benchmark results.
- Official name
- Mistral Large 4
- Nickname
- Le Chonk
- Model size
- One trillion parameters
- Terminal-Bench 4.0 score
- 41%, up from 28% in the preview
- Coding Agent Index score
- 48, the same as Codex GPT-6.1 Terra
- Mistral funding
- More than €6.6 billion raised since its founding in 2023
4 sources
Met een dikke kat zegt het Franse AI-bedrijf Mistral: kijk ook eens naar ons
Mistral’s Le Chonk is now a top open model for coding, ranking alongside Codex GPT-6.1 Terra
Large 4 : comment Mistral a rattrapé les IA chinoises en code
Mistral Says Its AI Model ‘Le Chonk’ Has Leapfrogged China in the Open-Weight Race
What people are saying
It’s a pivotal moment in the global AI leadership race
It’s clearly a dual-use technology
The cyber defense capabilities will enable enterprises and governments to defend themselves against threat actors that are jailbreaking closed models to perform cyberattacks
Mistral probeert hetzelfde type product te ontwikkelen als deze bedrijven