Skip to content

Mistral AI

Mistral Large 4 (preview)

Mistral Large 4 (preview) is a model with open weights from Mistral AI. On the Vals AI Legal Research Bench it reaches 31.73 percent all-pass, position 52 of 91, within a standard error of ±3.23 percentage points at $2.72 per task, as at 9 November 2026 [1]. It does not pass the compliance gate: Public preview: under Mistral's commercial terms (§4.2, §4.3) Mistral may use customer data and output of preview models to train models, and opt-out or zero data retention do not apply there (Commercial Terms effective 2026-09-25, read 2026-11-10). docs.mistral.ai lists Mistral Large 4 (26.10) as GA since 2026-11-05: see model mistral-large-4-ga.

  • Public preview
  • Open weights
  • 6 October 2026
PrudentScore
-No score: the gate is not open
Compliance gate
RejectedPublic preview: under Mistral's commercial terms (§4.2, §4.3) Mistral may use customer data and output of preview models to train models, and opt-out or zero data retention do not apply there (Commercial Terms effective 2026-09-25, read 2026-11-10). docs.mistral.ai lists Mistral Large 4 (26.10) as GA since 2026-11-05: see model mistral-large-4-ga
Cost per task
$2.72Vals AI, per completed task [1]

01

Compliance gate

Why

  • Public preview: under Mistral's commercial terms (§4.2, §4.3) Mistral may use customer data and output of preview models to train models, and opt-out or zero data retention do not apply there (Commercial Terms effective 2026-09-25, read 2026-11-10). docs.mistral.ai lists Mistral Large 4 (26.10) as GA since 2026-11-05: see model mistral-large-4-ga.

The gate only opens when all three requirements are covered: processing in the EU, a contracting party in the EU or EEA, and a contractual exclusion of training on customer data. The gate tests a route, not a model name: the same model can pass through one provider and fail through another. Where a provider publishes no figure, we give no estimate.

02

Routes and EU processing

The gate tests a route (processor, region and contract), not a model name. Per route: processing in the EU, the contracting party and training on customer data.

No route has been recorded in the register for this model yet.

03

Scores per benchmark

Every measurement with its source, measurement date and the day we read it. The position counts only the models measured on the same benchmark and version.

  • Legal Research BenchVersion 1 · Counts towards the PrudentScore
    31.73%± 3.2352 of 91

    Vals AI, Legal Research Bench, v1, 9 November 2026 · Source · Measured 9 November 2026 · Read 10 November 2026 · cost per task $2.72

  • 17.50%± 3.228 of 36

    Vals AI, Harvey LAB (Vals implementation), v1.1, 9 November 2026 · Source · Measured 9 November 2026 · Read 10 November 2026 · cost per task $2.55

  • Harvey LAB (Vals implementation)Version 1 · Counts towards the PrudentScore
    15.83%± 2.9610 of 80

    Vals AI, Harvey LAB (Vals implementation), v1, 18 October 2026 · Source · Measured 18 October 2026 · Read 10 November 2026 · cost per task $4.18

  • LegalBench (Vals leaderboard)Not measured
  • CaseLaw v2Not measured
  • TaxEval v2Not measured
  • 30.39%± 3.3133 of 82

    Vals AI, Tax Agent Bench, v1, 9 November 2026 · Source · Measured 9 November 2026 · Read 10 November 2026 · cost per task $1.39

  • CorpFin v2Not measured
  • 63.71%± 1.6620 of 26

    Vals AI, Immigration Bench, v1, 9 November 2026 · Source · Measured 9 November 2026 · Read 10 November 2026 · cost per task $0.15

  • Access to JusticeNot measured
  • Public Benefits BenchNot measured
  • 65.61%± 1.6329 of 36

    Vals AI, Vals Legal Index, v1, 9 November 2026 · Source · Measured 9 November 2026 · Read 10 November 2026 · cost per task $2.22

  • Vals IndexVersion 2.1
    48.66%± 1.1335 of 50

    Vals AI, Vals Index, v2.1, 18 October 2026 · Source · Measured 18 October 2026 · Read 10 November 2026 · cost per task $14.07

  • GPQA Diamond (Vals-run)Not measured
  • MMLU Pro (Vals-run)Not measured
  • Artificial Analysis Intelligence IndexVersion 4.3.2 · Counts towards the PrudentScore

    Artificial Analysis Intelligence Index v4.3.2: 38.4. Source: Artificial Analysis (artificialanalysis.ai), accessed 20 October 2026 [37].

    Artificial Analysis, Artificial Analysis Intelligence Index, v4.3.2, 20 October 2026 · Source · Measured 20 October 2026 · Read 20 October 2026

Source: Vals AI (Legal Research Bench, Harvey LAB (Vals implementation), Tax Agent Bench, Immigration Bench, …) · Licence: CC BY 4.0 · Adaptation: Prudai selected the models, translated the explanations, added the gate verdict and the PrudentScore and rounded figures to two decimals.

The PrudentScore is a composition by Prudai B.V. based on publicly published figures and does not represent the view or endorsement of Vals AI, Harvey or Artificial Analysis. Figures from Artificial Analysis appear in running text only, with attribution and a link, as that party's terms require [9].

04

Key facts

Developer
Mistral AI
Model family
Mistral
Release date
6 October 2026
Status
Public preview
Context window
512K tokens
Max output tokens
256K tokens
List price
$0.70 input · $2.10 outputper million tokens · Read 10 November 2026
Weights
Open

- = not yet recorded in the register.

05

Frequently asked questions about this model

Does Mistral Large 4 (preview) pass the compliance gate for legal work?

No. Public preview: under Mistral's commercial terms (§4.2, §4.3) Mistral may use customer data and output of preview models to train models, and opt-out or zero data retention do not apply there (Commercial Terms effective 2026-09-25, read 2026-11-10). docs.mistral.ai lists Mistral Large 4 (26.10) as GA since 2026-11-05: see model mistral-large-4-ga. The gate tests a route, not a model name; the sources are on this page.

What does Mistral Large 4 (preview) cost per task?

Vals computes $2.72 per completed task on the Legal Research Bench [1]. The list price is $0.70 per million input tokens and $2.10 per million output tokens.

How does Mistral Large 4 (preview) score on legal research?

31.73 percent all-pass on the Vals AI Legal Research Bench, position 52 of 91 measured models, as at 9 November 2026 [1]. That benchmark measures US law; an independent, published benchmark for Dutch legal work does not exist yet.

Related models

Claude and Anthropic, GPT and OpenAI, Gemini and Google, Grok and xAI, Kimi and Moonshot AI, Muse and Meta, Mistral, DeepSeek, GLM and Z.ai, Harvey, Vals AI and Artificial Analysis are trademarks of their respective owners. They are named for identification only and this implies no partnership, sponsorship or endorsement.

Brand marks appear for identification only and imply no partnership, sponsorship or endorsement.