All AI models
Open sourceVerified October 23, 2026

Mistral AI

Mistral Small 4

Mistral Small 4 is a compelling open-source all-rounder when you want text and image input, configurable reasoning, agentic coding support, and a low published API rate. It is especially attractive when a permissive license matters more than chasing the largest possible context window.

License
Apache 2.0
Context
256K tokens
Modalities
Text · Image
Parameters
119B total, 6.5B active

What the license means

Mistral released Small 4 under Apache 2.0. The permissive terms make it a useful candidate for customization, but production teams should still review the selected runtime, hosting, and security setup.

Open-weight, open source, and no-cost API access mean different things. Always review the linked terms and your own data, hosting, and compliance requirements before shipping.

API and operating cost

Input / 1M
$0.15
Output / 1M
$0.60

Official API price per 1M tokens for mistral-small-2603, as listed on October 23, 2026.

Check current provider price

Where it fits

  • Multimodal document analysis
  • Cost-aware coding agents
  • Customizable general assistants

Where it does not fit

  • One-million-token workloads
  • Single-machine serving without careful quantization and capacity testing

Local deployment reality

Mistral documents support through vLLM, llama.cpp, SGLang, and Transformers. Its official minimum deployment examples use multi-GPU hardware, so a smaller quantization is more realistic for an individual local setup.

vLLMllama.cppSGLangTransformers

Practical setup notes

  • Start with reasoning disabled for straightforward routing and extraction work.
  • Test image tokenization and long-document cost before committing to a multimodal production design.

Capability notes

Reasoning
Configurable reasoning effort
Tools and output
Agentic coding and common open inference runtimes

Limitations to plan around

  • The official minimum serving examples use substantial GPU hardware.
  • Context is 256K, not a 1M long-context offering.

Continue comparing

Related model decisions

See all reviewed models
Open source

Qwen3.6

Local coding assistance

Open model family

Gemma 4

Right-sized open-model experiments

Open-weight

DeepSeek V4

Cost-sensitive API workflows

Editorial note: This is a dated decision page, not a benchmark leaderboard or a substitute for legal, security, or infrastructure review. Model and pricing facts can change after October 23, 2026; use the linked primary sources before committing budget or customer data.Rechecked October 23, 2026: model ID (mistral-small-2603), license, parameter count, context window, and API price. Runtime and reasoning notes date from the August 11, 2026 audit.