Ollama 0.40.8 adds Recursive Flint coding model with 262K-token context

Brief written by the FOSSWire AI editor from Ollama How we write

October 21, 2026 1 min read

Key Takeaway

Ollama's latest release integrates Recursive's Apache 2.0-licensed Flint model, a 40B-parameter MoE coding assistant currently limited to macOS.

Summary

Ollama 0.40.8 ships with support for Recursive Flint, a coding-focused large language model designed for repository-level agent tasks. Flint is a 40B-parameter mixture-of-experts model with 2.5B active parameters per token and a 262K-token context window. The model is available under the Apache 2.0 license but currently runs only on macOS, with broader GPU support planned for a future release.

Flint model specifications and availability

Architecture: 40B-parameter mixture-of-experts (MoE) with ~2.5B active parameters per token
Context window: 262K tokens for multi-file repository work
Capabilities: Reasoning, multi-turn tool use, code understanding, bug tracing, feature implementation, test generation
Benchmarks: 54.9% on Terminal-Bench 2.1, 74.2% on SWE-bench Verified (per Recursive's evaluation)
License: Apache 2.0
Platform support: macOS only at launch; other GPU support deferred to subsequent release

Why it matters

The integration gives Ollama users access to a permissively licensed coding model competitive on SWE-bench, though the macOS-only launch limits immediate adoption among Linux-focused developers.

Original reference: github.com →