LMSYS trains Qwen3 coding agent with asynchronous RL on AMD MI355X
LMSYS Blog reports that its team trained Qwen3-30B-A3B-Instruct for coding-agent tasks using fully asynchronous agentic reinforcement learning on AMD Instinct MI355X, with Miles, SGLang, Megatron-LM, and Harbor. On a vendor-run, 100-task stratified subset of SWE-bench Verified, pass@1 rose from 7.3% to 27.5% after supervised fine-tuning and agentic RL. The system used separate inference and training nodes, capped trajectories at 64K tokens, and moved sandboxes to E2B cloud microVMs, supporting 384 concurrent trajectories and about 3,480 trajectories per hour.
The report provides concrete architecture and scaling details for asynchronous coding-agent training on AMD hardware, including context limits, sandbox concurrency, and measured throughput.
On this page
#Qwen3-30B-A3B-Instruct#LMSYS#AMD Instinct MI355X#Miles#SGLang#Megatron-LM#Harbor#E2B#SWE-bench Verified