Mixtral 8x22B logo

Mixtral 8x22B

Paid

Cheaper, Better, Faster, Stronger

5.0
Inputs: textOutputs: text
Type
Saas
Founded
2023
Company
Mistral AI

About Mixtral 8x22B

Mixtral 8x22B is a sparse Mixture-of-Experts (SMoE) language model that activates only 39B of its 141B total parameters per inference, delivering strong performance with high cost efficiency. Released under the Apache 2.0 open-source license, it supports a 64K token context window, native function calling, and constrained output modes for application development. The model achieves top benchmarks in reasoning, mathematics, and coding, outperforming many dense 70B models while being faster. It offers native fluency in English, French, Italian, German, and Spanish, making it suitable for multilingual tasks. The base model is available for fine-tuning, and an instructed version improves math performance (90.8% on GSM8K maj@8).

Key Features

Sparse Mixture-of-Experts with 39B active parameters out of 141B
64,000-token context window
Native function calling and constrained output mode
Multilingual in English, French, Italian, German, and Spanish
Open-source under Apache 2.0 license
Strong math and coding performance (90.8% GSM8K maj@8 on instructed version)
Optimized for reasoning tasks
Faster than dense 70B models with higher capability

Pros & Cons

Pros
  • Fully open-source with permissive Apache 2.0 license
  • Excellent cost efficiency due to sparse activation
  • Strong performance in reasoning, math, and coding benchmarks
  • Supports function calling natively, enabling complex workflows
  • Large 64K token context window for long-document tasks
  • Multilingual out-of-the-box (5 European languages)

Best For

Multilingual applications and content generationMathematical problem solving and code generationFine-tuning for domain-specific tasksBuilding scalable applications with function callingProcessing and reasoning over large documents (64K context)

Alternatives to Mixtral 8x22B

FAQ

What is Mixtral 8x22B?
Mixtral 8x22B is an open-source sparse Mixture-of-Experts language model developed by Mistral AI, using 39B active parameters out of 141B total. It is designed for high efficiency and strong performance in reasoning, math, coding, and multilingual tasks.
What license is Mixtral 8x22B released under?
It is released under the Apache 2.0 license, the most permissive open-source license, allowing unrestricted use, modification, and distribution.
What languages does Mixtral 8x22B support?
It natively supports English, French, Italian, German, and Spanish with strong multilingual performance.
What is the context window size?
Mixtral 8x22B has a 64,000-token context window, enabling precise information recall from large documents.
Does Mixtral 8x22B support function calling?
Yes, it natively supports function calling and constrained output modes, which facilitate application development and tech stack modernization.