Devstral 2: The best open-weights model for Cline just dropped
Mistral AI released Devstral 2 this week, and it immediately became one of the strongest open-source options for agentic coding. If you're running Cline, this matters.
Devstral 2 hits 72.2% on SWE-bench Verified with near parity with the best closed models while being up to 7x more cost-efficient than Claude Sonnet on real-world tasks. It's currently free during the launch period. The model family comes in two sizes: Devstral 2 (123B) and Devstral Small 2 (24B). Both support 256K context windows and are released under permissive open-source licenses.
Why this works for agentic coding
Devstral 2 excels at the multi-step, tool-using workflows that define modern coding agents. The tool-calling success rate matches the best closed models, which means fewer failed attempts and cleaner execution.
The 256K context window enables architecture-level reasoning across large codebases. Devstral understands framework dependencies, tracks state across multiple files, and can retry with corrections when something fails. This is exactly what you need for complex refactoring, bug fixes, or modernizing legacy systems.
Human evaluations show Devstral 2 beating DeepSeek V3.2 with a 42.8% win rate versus 28.6% loss rate. While Claude Sonnet 4.5 still leads in some areas, the gap between open and closed models keeps shrinking.
Model specs
Devstral 2 is a 123B-parameter dense transformer released under a modified MIT license. It's 5x smaller than DeepSeek V3.2 and 8x smaller than Kimi K2, proving compact models can compete with much larger alternatives.
Devstral Small 2 scores 68.0% on SWE-bench Verified despite being only 24B parameters. It's released under Apache 2.0, runs on consumer hardware including GeForce RTX, and supports image inputs for multimodal workflows.
Both models handle:
- Deep code understanding across entire codebases
- Multi-file orchestration with architectural context
- Framework dependency tracking and automatic retry logic
- Custom fine-tuning for specialized use cases
Deployment and pricing
API pricing after the free launch period:
- Devstral 2: $0.40 input / $2.00 output per million tokens
- Devstral Small 2: $0.10 input / $0.30 output per million tokens
Using Devstral 2 with Cline
Prerequisites:
- A Cline account (free to create)
- Cline installed in your IDE (available for VS Code, Cursor, Windsurf, and other VS Code forks)
Setup
- Open Cline
- Go to Settings
- Select Cline as an API provider
- Select mistralai/devstral-2512:free or any of the variants from the model dropdown.
- Done!
The model handles tool execution smoothly, understands complex instructions, and maintains context across long conversations. It's particularly strong at multi-file edits and systematic refactoring.
What this means for open source
Compact, efficient models that match larger competitors change the deployment equation. Devstral 2 runs on more accessible hardware while delivering performance that was confined to massive clusters just months ago.
The open-weights approach means you can fine-tune for your specific needs, run completely private infrastructure, or switch between local and cloud deployment based on your workflow.
Try it free during the launch period. If you're looking for a powerful open-source model that handles agentic workflows well, Devstral 2 is worth testing.
For questions or to share your experience with Devstral in Cline, join us on Discord at https://discord.gg/cline or Reddit at https://www.reddit.com/r/cline/.
