Skip to content

FastContext 1.0 4B RL

microsoft/FastContext-1.0-4B-RL
Open Source · chat · open-weights
GA Alert me on changes
Context
262.1K
Max output
Pricing
Open weights
Modalities
text
Released
14 Jun 2026
AI summary
● machine-written

Microsoft releases FastContext 1.0 4B RL, a 4B coding agent subagent

FastContext-1.0-4B-RL is a 4 billion parameter model developed by Microsoft for repository exploration in coding agents. It reduces token consumption by up to 60% and improves end-to-end resolution rates by up to 5.5% through parallel tool calls and focused context provision. The model supports 262K token context and uses reinforcement learning refinement based on Qwen3-4B-Instruct.

What's new
  • 4B parameter model optimized for repository exploration in coding agents
  • Supports 262K token context window for extended code navigation
  • Reduces main-agent token consumption by up to 60% on benchmark tasks
  • Improved end-to-end resolution rates up to +5.5% when integrated with main agents
  • Uses GRPO-based reinforcement learning for search and citation generation
Best for
Building or enhancing LLM-powered coding agents requiring efficient repository navigationReducing token consumption in multi-agent coding systemsFast code exploration and information extraction from large repositoriesImproving accuracy of main coding agents through focused context provision
Sources

Source: https://huggingface.co/microsoft/FastContext-1.0-4B-RL