Mistral Small 3: pricing, context and performance

Mistral Small 3 at a glance

Mistral Small 3 is a language model from Mistral AI, released on January 30, 2025. That makes it more than a year old: successors may have shipped since. Its weights are published: you can download it and run it on your own servers, or use one of the 6 API providers that list it on models.dev. It has no declared reasoning mode: it answers directly, without an extra billed thinking step.

We file it under open-weight models (68 models in this category on the site).

What do independent benchmarks say?

Epoch AI publishes 2 results for Mistral Small 3 on our basket's benchmarks: too few to compute a reliable capability index.

• GPQA Diamond (PhD-level science questions): 47.3% — the best tracked model, Gemini 4 Argon, reaches 96.4%
• OTIS Mock AIME (competition mathematics problems): 6.7% — the best tracked model, GPT-5.5 Pro, reaches 100%

What it costs

Mistral Small 3 costs $0.05 per million input tokens and $0.08 per million output tokens (median of 6 providers, checked on October 22, 2026).

To put it in perspective: reading 20,000 tokens (about thirty pages) and writing 2,000 costs roughly $0.001; a thousand of those tasks, $1.16.

That is 12.5 times cheaper than the median output price of open-weight models ($1).

To estimate your own budget, use the agent cost calculator.

Context and capabilities

Its context window reaches 32,768 tokens, roughly 49 pages of text at 500 words per page: best kept for short exchanges and brief documents. A single answer can run to 16,384 tokens.

It accepts as input: text. Its knowledge stops at October 31, 2023: beyond that, it needs web search or your documents.

It is not listed as able to call tools, so it is a poor fit for driving an agent that has to act on other software.

The weights are published on Hugging Face (mistralai/Mistral-Small-24B-Instruct-2501). See also our list of open-source models.

When to choose it

Given its category, Mistral Small 3 is mainly aimed at projects where you want control over your data or hosting, or want to adapt the model to your field.

Its very low cost (under half a cent for our sample task) means it can be called thousands of times a day.

Before adopting it, test it on your own examples: benchmarks measure standard tasks, not your job.

What could replace it?

In the same category, the models closest to Mistral Small 3 are:

• DeepSeek R1 0528 Qwen3 8B (DeepSeek, $0.09 per million output tokens)
• Ministral 3 3B 2512 (Mistral AI, $0.10 per million output tokens)
• Granite 4.0 Micro (IBM (Granite), $0.11 per million output tokens)
• LFM2-24B-A2B (Liquid AI, $0.12 per million output tokens)