Orange Cyberdefense has joined Mistral AI’s design-partner programme with early access to a cybersecurity-focused variant of Mistral Large 4. Announced on 29 October, the agreement will bring the model gradually into selected security services, beginning with penetration testing, technical audits and incident investigation.[1]
Those starting points put the model on both sides of a breach: helping specialists explore weaknesses before an attack, then analyse traces and reconstruct events afterwards. Orange says access controls, traceability and model security testing will accompany the work, with experts retaining responsibility for critical actions.[1]
Two routes into Large 4
Mistral’s public release of Large 4 weights gives organisations a way to run and adapt the general model. The specialised cybersecurity variant used by design partners has a different route: preview access subject to vetting and contractual safeguards. Mistral limits that testing to systems the user owns or is explicitly authorised to assess, in controlled environments.[2]
The distinction follows a programme outlined on 6 October, when Mistral launched the public API preview. It said security companies, vetted partners and state authorities would help test expanded cyber capabilities.[3]
Mistral says organisations can tailor Large 4 with their security data, policies and detection logic through Forge. For Orange, the potential advantage is therefore more specific than adding a chat interface: the model can be connected to the evidence and procedures its specialists use when assessing an incident.[2]
Where the work will run
Orange plans to use computing capacity in Mistral’s European data centres. That specifies where its cyber models will operate; the announcement does not provide a service launch date or measured reduction in investigation time.[1]
The Next Web reports that the partnership follows Orange Cyberdefense’s use of Qevlar AI in its security operations centres since April 2025. The financial terms of the new agreement were not disclosed. It adds a model-development relationship to a provider already using AI in operational security work.[4]
Orange says experience from these services will feed back into the model’s development.[1]
Sources & further reading
- Orange Cyberdefense becomes Mistral AI’s first European MSSP design partner Orange Cyberdefense · 29 Oct 2026 · Accessed 29 Oct 2026
- Mistral Large 4 (Le Chonk) is now open weight Mistral AI · 29 Oct 2026 · Accessed 29 Oct 2026
- Introducing Mistral Large 4 Mistral AI · 6 Oct 2026 · Accessed 29 Oct 2026
- Orange Cyberdefense joins Mistral design partner programme for Large 4 The Next Web · 29 Oct 2026 · Accessed 29 Oct 2026