vLLM says prime-rl runs trillion-scale agentic RL on 28 H200 nodes
vllm_project · x · 2026-07-23
vLLM Project says Prime Intellect’s prime-rl 0.6.0 is running trillion-scale agentic RL on vLLM, using FP8, wide expert parallelism, prefill/decode disaggregation, KV cache offloading, and vllm-router. The setup is used to train GLM-5 on SWE tasks with a 131k sequence length and sub-5-minute steps across 28 H200 nodes.
The post also points readers to a live vLLM Office Hours session featuring a deep dive on prime-rl’s performance and a broader update on vLLM 0.25.
More from Infra
- Anthropic's S-1 Reveals Over $600B in Infrastructure Commitments Across Google, Amazon, Microsoft — cedric_chee · 2026-10-26
- Anthropic S-1 reveals 48 cents of compute cost per dollar of Q2 revenue — nmasc_ · 2026-10-26
- Persistent AI joins UR network to route private AI inference with encrypted, IP-hidden traffic — 0xSammy · 2026-10-26
- Anthropic's disclosed compute commitments: $114.5B Google, $161.2B Broadcom top the list — BenBajarin · 2026-10-26
- SpaceX strikes $40B deal with Apollo Global to lease AI chips and data center hardware — Polymarket · 2026-10-26
- SpaceX AI revenue up 830%, Claude designs 318K antivenom proteins in 64 hours — PeterDiamandis · 2026-10-26