LANGUAGE MODEL Mistral AI Last updated:

Mistral Large 3 (v25.12)

Mistral Large 3 (v25.12) is an API model from Mistral AI. It’s positioned for vision + text tasks—work that benefits from iteration, not just one-shot answers.

Intelligence
Medium
Speed
Slow
61 tok/s output
Cost
Moderate
$0.50 in / $1.50 out
Context
256K
Up to 256,000 tokens
How are Intelligence, Speed & Cost bucketed?
Intelligence and Speed buckets are percentile ranks on Artificial Analysis. Cost buckets are fixed dollar thresholds keyed off output-token price ($/M out).
Intelligence
  • Top 1%≤ 1%
  • Top 5%≤ 5%
  • Top 10%≤ 10%
  • Good≤ 25%
  • Medium≤ 50%
  • Below avg> 50%
Speed
  • Top 1%≥ 345 tok/s
  • Top 5%≥ 237 tok/s
  • Top 10%≥ 196 tok/s
  • Good≥ 146 tok/s
  • Medium≥ 90 tok/s
  • Slow< 90 tok/s
Cost
  • Freeopen weights · self-host
  • Low< $1 / M out
  • Moderate$1–5 / M out
  • High≥ $5 / M out

Core Capabilities

Long Documents
Handles entire codebases, books, and multi-doc RAG.
Multimodal
Combines text, vision, and audio in one model.
Vision
Understands images, scenes, and visual context.

Context Window

256k tokens
≈ 197+ pages
4k Chat 聊天
32k Long docs 长文档
128k Books 整本书
400k Multi-doc 多文档
1M Codebase 整个代码库
10M
256k

Availability

API
Available
Product / App
Not available
Open Source
Not released
Enterprise
Contact sales

Pricing Model

Free / self-host
Open weights — pay only for compute.
Self-host

Capability / Performance

Where this model sits relative to the middle 60% of models in the tree. All scores are 0–10 (higher is better).

Lower 20% Upper 80% This model
Reasoning
AA Intelligence Index · scaled to 10
1.7
5.6
3.3
Coding
SciCode · scaled to 10
1.8
4.3
3.6
Agentic tasks
Terminal-Bench Hard · scaled to 10
0.2
3.6
1.6
Context / memory
Context window size · log-scaled
6.0
9.0
7.0
Cost efficiency
Input price ($/M tokens) · cheaper scores higher
6.2
10.0
10.0
Lower 20% 20th percentile — 20% of models score below this This model Where the current model lands Upper 80% 80th percentile — only 20% of models score above this Percentile boundaries are computed across every model in the tree that reports the underlying benchmark for each capability.

What it feels like

  • 675B sparse MoE / ~41B active — Mistral's biggest model and first with 256K context
  • Beats Kimi-K2 and DeepSeek-3.1 on Mistral's headline benchmarks; 1418 LMSYS Elo at release
  • MMLU 88.7% and HumanEval 92.3% — frontier-tier on coding, top of the open-weight pack
  • Solid but not best-in-class on hardest math contests — 'more than adequate' for business analytics
  • Cheap: $0.50/M in, $1.50/M out — among the most affordable frontier-tier APIs
  • Slower than average at ~50 tok/s — pick this for quality/price, not for chat speed

Best use cases

  • Vision + text workflows that benefit from iteration (image input is first-class)
  • Long-document analysis up to 256K tokens at affordable open-weight pricing
  • European customers needing GDPR-friendly hosting and EU-anchored vendor
  • Self-hosted production once weights ship — TechCrunch confirmed open-weight release

Tools to try

Not ideal for

  • Latency-sensitive interactive chat — throughput is below average
  • Hardest reasoning leaderboards — Claude Opus 4.5, GPT-5, DeepSeek V4 score higher
  • Pure-text deployments where smaller Mistral models would be cheaper