Mistral Medium 3.5

Overview

Mistral Medium 3.5 is a 128-billion-parameter dense instruction-following model from Mistral. It accepts text, images and files, and returns text. The model is positioned for agentic workflows, coding and complex tasks, with tool support and reasoning enabled.

Its 262,144-token context window gives teams room to supply long inputs, while the maximum output is 209,715 tokens. Those limits describe the model's available context and output capacity; whether they suit a particular workflow depends on the size and shape of its prompts and responses.

Medium 3.5 belongs to the Mistral models lineup. Its image input also places it among vision language models, and its tool support and reasoning capability are relevant to teams comparing models with tool calling and reasoning models.

Key features

Text, images and files

The model takes text, image and file inputs but produces text output. That combination makes it applicable to workflows that need to bring visual or file-based material into a text-generating process, alongside ordinary text prompts.

Long context and tool support

A 262,144-token context window is a central specification for teams working with substantial inputs or extended exchanges. The model supports tools, which is relevant to agentic workflows that incorporate tool use. The available facts establish tool support, but do not specify particular tools or integrations.

Reasoning and coding

Reasoning is enabled, and the model is designed for coding and complex tasks as well as agentic workflows. These are intended use areas rather than a guarantee of performance on a team's specific codebase or task.

Pricing

Input costs $1.5 USD per 1M tokens. Output costs $7.5 USD per 1M tokens. Output tokens are priced at five times the input rate, so teams should account for both sides of usage when estimating spend.

The model's maximum output is 209,715 tokens, but that capacity is separate from its per-token price: a response's cost depends on the number of tokens actually generated.

Platforms

The listed model identifier is mistralai/mistral-medium-3-5. The available platform facts identify text, image and file inputs, text outputs, tool support, and the context and output limits. They do not specify deployment options, SDKs or integrations, so teams should confirm those details against their own platform requirements.

Who it's for

Mistral Medium 3.5 is a candidate for teams evaluating an instruction-following model for coding, complex tasks or agentic workflows. Its image and file inputs may fit stacks where those materials need to accompany text, while tool support is relevant where workflows use tools. The long context window may matter when prompts or working material are unusually large.

It is also worth considering for teams that can budget separately for input and output tokens and that need a text-generating model rather than multimodal output. The listed output modality is text.

Pros and cons

• Pros: supports text, image and file inputs; includes tool support and reasoning; offers a 262,144-token context window; supports outputs up to 209,715 tokens.
• Cons: output tokens cost more than input tokens; output is text-only; the listed facts do not identify specific tool integrations or deployment choices.

Alternatives

For another listed form of this model, compare Mistral Medium 3.5 (batch). Other Mistral options include Mistral Small 4 and Devstral 2 2512. Smaller-model names in the supplied alternatives include Ministral 3 14B 2512, Ministral 3 8B 2512 and Ministral 3 3B 2512. Batch-listed options also include Mistral Small 4 (batch) and Ministral 3 8B 2512 (batch). The available facts do not provide their specifications or prices for direct comparison.

Browse newest models or longest-context models for broader model lists.

Verdict

Mistral Medium 3.5 combines a broad input set, tool support, reasoning and a large context window in a text-output model aimed at coding, agentic workflows and complex tasks. The main pricing consideration is its $7.5 USD per 1M tokens output rate against $1.5 USD per 1M tokens for input. Teams can weigh that cost structure and the stated capabilities against the requirements of their existing stack; the listed facts do not establish particular integrations or comparative performance.