MistralBase plan

Devstral 2

Mistral AI's Devstral coding model via the Vercel AI Gateway: agentic software engineering, multi-file edits, and tool use at a 256k context window. Available at the base tier. Uses 3 base requests per send before length multipliers. Does not support web search or image input.

You always get the exact model you pick — we never silently route you to another.

About Devstral 2

Devstral 2 is Mistral's coding-focused model, tuned for agentic software engineering rather than general chat, and on just4o.chat it is served through the Vercel AI Gateway as mistral/devstral-2. It targets the things real coding work demands: multi-file edits, navigating and reasoning about a codebase, and using tools in a loop to get a task done. A 256k-token context window means it can hold a sizable chunk of a repository at once, and function calling lets it drive editors, test runners, and other tooling. On just4o.chat it is a base-tier model on every plan, billed at 3 base requests per send before length multipliers, at a provider price of $0.40 per million input and $2.00 per million output tokens. It is text-only with no native web search, so it is built for the editor, not for research. For developers who want a capable, affordable coding agent that can work across files, Devstral 2 is a strong fit; for lighter or higher-volume coding, its smaller sibling Devstral Small 2 costs even less.

Best for

  • Agentic software engineering — multi-step coding tasks driven by tool calls
  • Multi-file edits and refactors across a codebase
  • Code generation, debugging, and root-cause analysis
  • Test scaffolding and iterative fix-and-verify loops
  • Working with large repositories thanks to the 256k context window

Specifications

ProviderMistral
Released2025-12
IntelligenceHigh
SpeedFast
Context window256,000 tokens
Knowledge cutoffNot officially published
Input price$0.40 / 1M tokens
Output price$2.00 / 1M tokens
Request cost3 base requests
Plan tierBase
InputText
OutputText
Features3 base requests per send before length multipliers, Function calling supported, Optimized for agentic coding workflows, Served through the Vercel AI Gateway (mistral/devstral-2), Mistral applies zero data retention to gateway traffic
Model IDdevstral-2

Frequently asked questions

Agentic coding — multi-file edits, codebase reasoning, and tool use — rather than general conversation. It is Mistral's developer-focused model.

Through the Vercel AI Gateway's OpenAI-compatible endpoint under the mistral/devstral-2 route. Mistral applies zero data retention to gateway traffic.

$0.40 per million input tokens and $2.00 per million output tokens at provider reference pricing. On just4o.chat it is a base model that uses 3 base requests per send before length multipliers.

No. Devstral 2 is text-only and has no native web search; it is designed for coding tasks in the editor.

Devstral 2 is the larger, more capable coding model; Devstral Small 2 is cheaper and faster for lighter or higher-volume work. Both share the 256k context window and function calling.

Related models