DOOMBENCH.COM
Mistral AI 🇫🇷 · Mistral Large

Mistral Large 4

Mistral Large 4 is a general-purpose multimodal model in a hosted public API preview. Current commercial terms reserve model rights and restrict access to proprietary components. The announced weight release remains a future plan; Closed describes the current preview, not the promised later artifacts.

DOOM SCORE78.2out of 100model risk profile, not the overall index
0 comments · 0 votesOpen discussion

Public discussion is readable by everyone. Sign in to comment, reply, or vote.

No comments yet. Start the discussion.

Send detailed or private feedback →
CURRENT ASSESSMENT · REVISION 2

Why this model scores 78.2

Dimensions remain unchanged: capability 87 reflects broad evaluated coding, cyber and multimodal work; autonomy 80 reflects agent workflows with residual failures; deployment 68 reflects hosted preview access, not released weights. Misuse 82 reflects dual-use cyber capability, while control difficulty 68 balances reported safeguards with unverified operational control. Current proprietary preview terms resolve availability without treating planned weights as released.

Capability87
Autonomy80
Deployment68
Misuse potential82
Control difficulty68
WATCH AND EXPLORE

Reviewed videos from Artificial Intelligence Videos. These links add context and never change a DoomBench score.

Thumbnail for Mistral's New AI Is WAY Better Than I Expected...
14:21

Mistral's New AI Is WAY Better Than I Expected...

AI Copium examines Mistral Large 4's selected benchmarks, GPU constraints, cybersecurity refusals, vision results and reinforcement learning.

Examines the exact Mistral Large 4 preview’s cyber refusal behavior, benchmark comparisons and compute constraints, clarifying consequential capability and misuse interpretations.
Thumbnail for Mistral Large 4: Open Weights and Harness Tradeoffs
21:29

Mistral Large 4: Open Weights and Harness Tradeoffs

Matthew Berman reviews Mistral Large 4's reported benchmarks and control benefits, then tests the model in a coding harness and explains integration problems.

Direct review of the exact Mistral Large 4 preview separates reported benchmarks and prospective weight control from observed coding-harness integration failures.
Thumbnail for Mistral Is BACK - Mistral Large 4 First Test (Le Chonk!)
38:50

Mistral Is BACK - Mistral Large 4 First Test (Le Chonk!)

Bijan Bowen tests Mistral Large 4 across browser apps, games and image-guided coding, finding promising progress alongside spatial and reliability limits.

Tests the exact Mistral Large 4 preview across coding and visual tasks, documenting concrete task-completion, spatial and harness-dependent reliability limits.
Thumbnail for Mistral Large 4 Tested on Eight Real Coding Tasks
9:14

Mistral Large 4 Tested on Eight Real Coding Tasks

AICodeKing tests Mistral Large 4 on eight coding tasks, separating working browser output from unfinished attempts and repairs.

Tests the exact Mistral Large 4 preview on eight coding tasks and identifies working, incomplete and repaired outcomes relevant to agentic task reliability.
MODEL-ATTRIBUTED EVIDENCE

News tied to Mistral Large 4

The model score of 78.2 rates this model's risk profile. The overall Doom Index of 67.9 measures the complete temporally weighted evidence record. These values answer different questions.

NET MODEL-ATTRIBUTED INDEX CONTRIBUTION+0.18

Each article's current Doom Index contribution is divided equally among the exact models named on that article. This prevents multi-model evidence from being claimed in full on several model pages. Model risk scores use a bounded temporal offset around their technical profile, but never feed back into the overall index.

Capability TOWARD

Mistral releases a cyber-capable Large 4 preview ahead of planned weight release

Mistral opened the Large 4 preview API on 6 October. Its launch reports strong cyber and agentic benchmarks and gives vetted red-team partners reduced moderation and expanded cyber access. Public weights are planned for later in October, not yet verified as released. Benchmark findings do not establish a real-world breach or job displacement.

Full item contribution
+0.18
Mistral Large 4 equal share
+0.18
Read assessment →
AUDIT TRAIL

Model score history

  1. R2
    Doom Score 78.2

    Availability only: October 6 hosted preview plus https://legal.mistral.ai/terms/commercial-terms-of-service/ sections 2.2, 4.3 and 5.1 establish restricted proprietary access. Weight release is still future; dimensions unchanged. Exact-version evidence chronology replayed under temporal-monthly-pressure-v4.

    12 Oct 2026
  2. R1
    Doom Score 78.2

    First exact-version assessment from the dated October 6 preview launch; weights remain a future plan, so no unsupported Open or Closed artifact classification.

    07 Oct 2026

Share this page

DoomBench social sharing card for Mistral Large 4.

  1. Mistral Large 4 by Mistral AI has a DoomBench model risk score of 78.2 out of 100, based on five transparent version-specific dimensions rather than the overall index.

  2. Mistral Large 4's highest current DoomBench dimension is capability at 87.0 out of 100; the profile publishes every component score and its editorial rationale.

  3. DoomBench links 1 source-backed evidence item to Mistral Large 4, while keeping the model's risk profile separate from each item's contribution to the live Doom Index.

    https://www.doombench.com/models/mistral-ai-mistral-large-4