DeepSeek-V4.1-Flash Hits 63x Throughput on Vera Rubin, With AI Agent Doing the Optimization
woosuk_k · x · 2026-11-17
The vLLM team published first DeepSeek-V4.1-Flash results on NVIDIA Vera Rubin NVL72, running Dynamo + vLLM on SemiAnalysis's AgentX benchmark. Versus GB300 NVL72 on the same stack, per-chip total token throughput is 63.4x higher at 400 tok/s/user and 93.3x at 600 tok/s/user.
Optimizer @liujuntian notes that as throughput rose, prefix cache hits dropped because duplicated Engram tables were eating KV cache memory. He tasked an AI agent with the fix; it shared Engram tables across the rack's GPU HBM via NVLink symmetric memory, with even lower lookup overhead. He says the agent could discover, implement and ship optimizations on brand-new hardware with only a few human hints.
Related event: vLLM Reveals DeepSeek V4.1 Flash Results on Vera Rubin NVL72(3 posts)→
More from coding & agent
- CMU paper: planning agents should show plans, not ask — acceptance jumps from 12% to 55% — rohanpaul_ai · 2026-11-17
- Boring Marketer open-sources 32 marketing agent skills and launches forward-deployed service — boringmarketer · 2026-11-17
- Pybot experiment write-up coming: how the autonomous GitHub bot actually ran — chrisalbon · 2026-11-17
- Running a company in 2031: a meme of one human in a meeting with nine Claude agents — andrew_n_carr · 2026-11-17
- Perplexity to upgrade its Slack agent Computer into a persistent assignable Teammate — AravSrinivas · 2026-11-17
- One chart mapping voice AI: S2S vs cascaded, half- vs full-duplex — rdesh26 · 2026-11-17