Mistral AI
Updated May 2026
Mistral
Europe's open-weight champion
Founded in 2023 by ex-DeepMind and Meta researchers in Paris, Mistral became Europe's AI national-champion by betting on an open-weights-first approach. They ship frontier-competitive models in 10-40B active parameters using mixture-of-experts, then offer API access for convenience. Their code model Codestral is a developer favorite, and Ministral 3B/8B are aimed at phones and edge devices.
Variants
5
Best at
open weights at frontier-adjacent quality
Max context
256K
tokens
Pricing
Free tier
$0
per limited API calls + Codestral free
Large 3 API
$2 in / $6 out
per million tokens
Medium 3 API
$0.40 in / $2 out
per million tokens
Le Chat Pro
$14.99
per month
Variants
Sort the table by context window or cost to find the right variant. Click any version below for a battle card with ranks, pricing notes, and official links.
| Modalities | ||||
|---|---|---|---|---|
Mistral Large 3 mistral-large-3 | 256K | varies / varies | 2025 | textvision |
Codestral codestral | 256K | varies / varies | 2024 | code |
Mistral Medium 3 mistral-medium-3 | 128K | varies / varies | 2025 | text |
Ministral 8B ministral-8b | 128K | varies / varies | 2024 | text |
Mixtral 8x22B mixtral-8x22b | 64K | varies / varies | 2024 | text |
Battle card
Context rank
#1
within Mistral
Capability rank
#1
modalities + reasoning
Weights
Open
self-hostable if licensed
Best fights to pick
- European-sovereign deployments
- multilingual reasoning
- enterprise fine-tuning
Rankings are Tendril directory ranks, computed from the model data shown here. Public benchmark leaderboards change often, so official docs and current benchmark pages should be checked before buying or deploying.
Learn
Lessons about this model
Structured lessons that cover Mistral directly or put it in context alongside its rivals.
Check yourself
Quizzes
Short, mixed-difficulty quiz sets on Mistral and its model family.
Open-Weight Families: Llama, Mistral, Qwen, DeepSeek, Gemma
7 questions
The open ecosystem that shook the industry.
Start quiz →Hands-on
Try these prompts
Ready-made prompts that show Mistral at its best. Use them in your own AI workspace, then compare the output with what you learned in Tendril.
Multilingual translation with Mistral Large
BuildersMistral's European roots show in multilingual work — test with a mid-resource language.
Translate this product description into French, German, Spanish, and Polish. Keep the friendly tone and technical terms accurate. 'Our noise-cancelling headphones use adaptive AI to learn your environment and cut commute noise by 40 decibels.'
Codestral fill-in-the-middle
BuildersTry Codestral's FIM mode — it's optimized for completing code with context on both sides.
Complete the middle of this function:
def normalize_email(email: str) -> str:
# Lowercase, strip whitespace, and for gmail addresses,
# remove dots from the local part and everything after '+'
<FILL HERE>
return emailEdge deployment feasibility check
CreatorsMistral Small is designed for phones and laptops — prompt it about its own resource requirements.
I want to run you on a MacBook Air M2 with 8GB of RAM. What quantization do you recommend, what tokens/second should I expect, and where will quality drop noticeably?
Print & keep
Printable reference
One-page summaries and flowcharts — great for desks, classrooms, or study sessions.
Go deeper
Official resources
Straight from the lab — docs, API references, and the chat surfaces you can try today.
Strengths
- open weights at frontier-adjacent quality
- excellent multilingual (especially French/German)
- fast inference on MoE architecture
Limits
- behind GPT/Claude on reasoning benchmarks
- smaller ecosystem than Llama
- Medium 3 is closed