Mistral: Models Intelligence, Performance & Price
This analysis is intended to support you in choosing the best model provided by Mistral for your use-case.
Most Intelligent
Mistral Large 4 (high)Mistral Large 4 PreviewMistral Large 4 (medium)Mistral Medium 3.5Mistral Small 4Intelligence Index
Total 11 models
Lowest Time per Task
Mistral Small 4Mistral Medium 3.5Mistral Large 4 (medium)Mistral Large 4 (high)Mistral Large 4 PreviewTime per Intelligence Index task (minutes)
Total 11 models
Lowest Price
Ministral 3 3BMinistral 3 8BMinistral 3 14BMistral Small 4Mistral Small 4 (non-reasoning)Blended price (per 1M tokens)
Total 11 models
Mistral offers 11 models, each with different intelligence, performance, and pricing characteristics. Below is a comparison of the key metrics across models.
- For intelligence, the top models on Mistral are Mistral Large 4 (high) (26), Mistral Large 4 Preview (25), and Mistral Large 4 (medium) (23).
- For time per task, Mistral Small 4 (1.6 min), Mistral Medium 3.5 (4.9 min), and Mistral Large 4 (medium) (8.9 min) finish an Intelligence Index task fastest on Mistral. Time per task varies significantly across models, with the slowest taking 12.2x longer than the quickest.
- For latency, Ministral 3 3B (1.51s), Mistral Small 4 (non-reasoning) (1.77s), and Ministral 3 8B (2.18s) offer the lowest time to first answer token.
- For pricing, Ministral 3 3B ($0.04), Ministral 3 8B ($0.06), and Ministral 3 14B ($0.07) offer the lowest blended prices per 1M tokens. Prices vary up to 2.7x across models.
- For context window size, GLM-5.2 (max) (1M), Mistral Large 4 (high) (1M), and Mistral Large 4 (medium) (1M) support the largest context windows on Mistral.
Intelligence
Artificial Analysis Intelligence Index
Evaluations
Agentic knowledge work, Elo
Knowledge
1 - hallucination rate
Agentic SaaS workflows
Agentic scientific research workflows in a terminal
Agentic software engineering in open-source repositories
Reasoning & knowledge
Professional document reasoning, All-pass
Agentic coding & terminal use
Physics reasoning
Quantitative analysis on spreadsheets & documents
Legal agentic work, Hallucination-Gated All-Pass Rate
Agentic business operations
Agentic real-world work tasks, Elo
Agentic interactive 3D experiences, Elo
Product and brand videos designed in code, Elo
Kubernetes incident root-cause analysis
Medical long context reasoning
Visual reasoning
Rich, varied, well-crafted language
Well-structured, scannable, answer-first
Truthful, acknowledges uncertainty, non-sycophantic
Use of filler and generic language
Intelligence Index vs. Price
Pricing
Cost per Intelligence Index Task
Intelligence Index vs. Price
Speed
Average of the single-request and multi-turn agentic workloads measured
Time per Intelligence Index Task
Quality
Context Window
Latency
Average of the single-request and multi-turn agentic workloads measured
Time to First Answer Token
Reliability
Share of requests the endpoint failed, across all the single-request and multi-turn agentic workloads measured
Endpoint Error Rate
Caching
Cache Hit Rate
Cost per Task vs. Cache Hit Rate
Further Analysis | ||||||||||
|---|---|---|---|---|---|---|---|---|---|---|
Mistral Large 4 (high) | 26 | $2.06 | 12.4 | 30.99 | 0.0% | 1M | Open | |||
Mistral Large 4 Preview | 25 | $4.21 | 19.5 | 38.33 | 0.0% | 524k | Proprietary | |||
Mistral Large 4 (medium) | 23 | $1.41 | 9.0 | 20.29 | 0.0% | 1M | Open | |||
Mistral Medium 3.5 | 5 | $1.04 | 4.9 | 27.01 | 0.0% | 262k | Open | |||
Mistral Small 4 | 4 | $0.04 | 1.6 | 12.61 | 0.9% | 256k | Open | |||
Mistral Small 4 (non-reasoning) | -- | -- | -- | 1.76 | 0.8% | 262k | Open | |||
Mistral Large 3 | -- | -- | -- | 3.26 | 2.7% | 262k | Open | |||
Ministral 3 3B | -- | -- | -- | 1.51 | 10.5% | 131k | Open | |||
Ministral 3 14B | -- | -- | -- | 2.53 | 3.6% | 262k | Open | |||
Ministral 3 8B | -- | -- | -- | 2.18 | 1.9% | 262k | Open | |||
 GLM-5.2 (max) | -- | -- | -- | 8.99 | 0.0% | 1M | Open | |||
Key definitions
Frequently Asked Questions
Common questions about Mistral
Mistral offers 11 models that we track: Mistral Large 4 (high), Mistral Large 4 Preview, Mistral Large 4 (medium), Mistral Medium 3.5, Mistral Small 4, Mistral Small 4 (non-reasoning), Mistral Large 3, Ministral 3 3B, Ministral 3 14B, Ministral 3 8B, and GLM-5.2 (max).
The most intelligent model available on Mistral is Mistral Large 4 (high) with an Intelligence Index score of 26.
The model with the lowest time per task on Mistral is Mistral Small 4 at 1.6 minutes per Intelligence Index task. Time per task is the output tokens per task divided by end-to-end output speed, so lower means a task finishes sooner.
The model with the lowest time to first answer token on Mistral is Ministral 3 3B at 1.51s. Lower latency means faster initial response time.
The most affordable model on Mistral by blended price is Ministral 3 3B at $0.037 per 1M tokens (7:2:1 cache hit/input/output ratio).
Prices on Mistral vary up to 31x across models, from $0.037 per 1M tokens for Ministral 3 3B to $1.16 per 1M tokens for Mistral Medium 3.5.
Yes, Mistral offers an OpenAI-compatible API, making it easy to switch from OpenAI or use existing OpenAI SDK integrations.
Yes, all 11 models on Mistral support JSON mode for structured output.
Yes, all 11 models on Mistral support function calling (tool use).
Yes, Mistral offers 6 reasoning models: Mistral Large 4 (high), Mistral Large 4 Preview, Mistral Large 4 (medium), Mistral Medium 3.5, Mistral Small 4, and GLM-5.2 (max). Reasoning models use extended thinking to work through complex problems before providing an answer.
Yes, 10 of 11 models on Mistral are open weight models: Mistral Large 4 (high), Mistral Large 4 (medium), Mistral Medium 3.5, Mistral Small 4, Mistral Small 4 (non-reasoning), Mistral Large 3, Ministral 3 3B, Ministral 3 14B, Ministral 3 8B, and GLM-5.2 (max).
Yes, provider performance can vary over time due to infrastructure changes, load balancing, and updates. We continuously benchmark all providers and display historical performance trends in the "Over Time" charts.
When choosing a model on Mistral, consider: intelligence (for quality-sensitive tasks), time per task (how quickly the model completes an Intelligence Index task), latency (for interactive applications requiring quick first responses), pricing (for cost-sensitive workloads), and features like context window size, JSON mode, or function calling support.