FireFlash releases recipe for 300B agentic RL model trained in 5 days
Researchers introduced FireFlash, a 300B-scale model for long-horizon agentic tasks, scoring 46.2% on Terminal-Bench 4.0 — on par with Fable 5 and GPT-6 Sol while costing over 30x less per task. It was trained in under 5 days on just 64 NVIDIA B300 GPUs, and the team says it will release the full recipe and open weights.