FastContext 1.0 4B RL
microsoft/FastContext-1.0-4B-RL
Open Source · chat · open-weights
Unavailable
Alert me on changes
Not currently available.
This model's weights can't currently be fetched from Hugging Face, so it isn't usable right now.
We re-check every few minutes and will alert subscribers the moment it returns.
Context
262.1K
Max output
—
Weights
Unavailable
API $/1M
—
Modalities
text
Released
14 Jun 2026
License: mit · microsoft/FastContext-1.0-4B-RL
AI summary
● machine-written
Microsoft releases FastContext-1.0-4B-RL, a 4B repository-exploration subagent
FastContext-1.0-4B-RL is a 4 billion parameter model designed to optimize LLM coding agents by efficiently exploring code repositories and returning compact file paths and line ranges. The model supports up to 262K token context length and was refined with reinforcement learning. It integrates with main coding agents to reduce token consumption by up to 60% while improving end-to-end resolution rates by up to 5.5%.
What's new
- Supports parallel READ, GLOB, and GREP tool calls for efficient repository exploration
- Context length of 262K tokens with reinforcement learning refinement
- Reduces main-agent token consumption by up to 60.3% on SWE-QA benchmarks
- Outperforms larger 30B-SFT models in some scenarios while using fewer tokens
Best for
Building LLM-powered coding agents that need efficient code repository navigationReducing token consumption in main coding agentsFast and precise code exploration scenarios
Source: https://huggingface.co/microsoft/FastContext-1.0-4B-RL