prime-rl 0.10.0 releases with IPO default and unified evals
Prime Intellect released prime-rl 0.10.0 and verifiers 0.4.0, making the new INTELLECT Policy Optimization (IPO) loss the default for asynchronous reinforcement learning. The release adds optimized model support, including DeepSeek V4 Flash, Qwen3.8 Flash Next and Nemotron-H, and consolidates evaluation into prime-rl. Prime Intellect reports that prime-rl was 1.15 times faster than torchtitan in its Qwen3-235B-A22B supervised fine-tuning comparison, though the benchmark details come from the vendor.
The release combines a new asynchronous RL loss, performance-oriented training kernels and unified evaluation tooling, making it directly relevant to practitioners building large-model reinforcement-learning workflows.
prime-rl 0.10.0: faster training, our new RL loss IPO, and one unified eval stack
Today we're releasing prime-rl v0.10.0 and verifiers v0.4.0. We added our new RL algorithm IPO as the default, tailored for async RL stability, as well as performance improvements and new model support.
Training throughput
The trainer's defaults favor memory, so large models fit on few GPUs. This release also documents the other end: a max-throughput recipe that keeps everything on GPU and compiles with fullgraph.
#Prime Intellect#prime-rl#verifiers#INTELLECT Policy Optimization#DeepSeek V4 Flash#Qwen3.8 Flash Next#Nemotron-H#GLM-5.3#Qwen3-235B-A22B#Qwen-VL#Kimi K2.5