Mistral Introduces Free Tier for Developers to Experiment with AI Models

Mistral AI recently unveiled a new free tier that empowers developers to fine-tune and create test applications using the startup’s advanced AI models, as detailed in a blog post on Tuesday. In addition to this offering, Mistral has significantly reduced the prices for accessing its AI models via API endpoints and has integrated image processing features into its consumer AI chatbot, le Chat.

There is a rising trend among AI model providers to deliver enhanced capabilities at lower costs. The Paris-based startup, now valued at $6 billion, announced various updates this Tuesday to captivate developers. As advanced large language models (LLMs) continue to become more readily available, Mistral aims to remain competitive against the dropping prices being offered by major players like OpenAI, Google, and Anthropic.

Both OpenAI and Google also provide free tiers that allow developers to experiment with their AI models, though these tiers come with lower rate limits compared to their premium options.

Mistral's new free tier is accessible through its API platform, la Plateforme, which the company asserts will enable developers to experiment, assess, and prototype with its AI models at no cost. However, to deploy Mistral’s models in a production environment, developers may need to upgrade to the paid commercial tier that offers higher rate limits.

Why does Mistral, which positions itself as providing “open” AI models, need to introduce a free tier for developers? While users can download Mistral's AI models under the Apache 2.0 license, establishing the necessary infrastructure to host these models can be complex. La Plateforme simplifies this process for developers, encouraging them to build on Mistral’s platform, where they can later be offered upgrades for higher usage limits.

Mistral also slashed prices for accessing its models—Mistral NeMo, Mistral Small, and Codestral—by over 50%, and reduced the price of Mistral Large by 33%.

Additionally, Mistral's first multimodal model, Pixtral 12B, which processes both images and text, is now available on le Chat. The new AI model, announced last week, is designed to enable users to scan, analyze, and search image files seamlessly.