Ollama has released version v0.40.8, introducing support for the Flint coding model. Flint, developed by Recursive, is a 40B-parameter mixture-of-experts model designed for complex coding tasks, featuring reasoning capabilities, multi-turn tool use, and a 262K-token context window. The release also includes enhancements to MLX attention for learned sinks and device causal masks, with Flint specifically opting into this new behavior. Flint is currently supported on macOS, with broader GPU support planned for a future release. AI
IMPACT Enhances coding agent capabilities with a large context window and specialized reasoning for complex tasks.
RANK_REASON This is a software release for a specific tool (Ollama) adding support for a model, not a frontier model release from a major lab.
- 262K-token
- Apache Software License 2.0
- macOS
- mlx
- Ollama
- SWE-bench Verified
- Terminal-Bench 2.1
- v0.40.8
- v0.40.8-rc0
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →