VarwoMistral Large 4: Access, Weights and Pricing
Check Mistral Large 4 access, announced API pricing and open-weight licence.
Mistral Large 4, also called Le Chonk, is Mistral’s multimodal model. Mistral describes a trillion total parameters and offers an API as well as downloadable weights. Total parameter count alone does not determine the resources required or the quality on your task.
API or self-hosted weights?
Mistral links to Studio, its API, the model card and the weights in its release. API users should confirm the actual model identifier, access and current billing in their account. Self-hosters should inspect the model card, serving requirements and licence before planning hardware.
As checked November 4, 2026, the release lists $0.70 per million input tokens, $0.07 cached input and $2.10 output. These are provider-listed rates, not a quote for another provider or a guarantee of future pricing. The release describes a Mistral Ecosystem Open-Weight licence and a separate commercial licence condition for model-as-a-service businesses. Read the actual licence and verify whether it covers your intended use.
Compare results on your own tasks
Mistral reports coding and security benchmarks in its launch materials. These are provider claims, not independent results from Varwo. Compare representative inputs, output quality, latency, corrections and total cost under matching conditions. Varwo reviewed documents but did not run the model or reproduce the benchmarks.
What to record before a comparison
Keep the model identifier, reasoning mode, task, input and output tokens, and the date of the quoted rate together. API access and self-hosted weights have different infrastructure and operating costs. Published leaderboards can suggest a test, but they cannot establish your own quality, latency or total cost.