7 results for “GB200”
NVIDIA reports its Transformer Engine with JAX lifts DeepSeek-V3 671B throughput to 1,068 TFLOPS/GPU on GB200. This dropless MoE stack makes large-scale training practical without padding.
September 2026 watch through 14 Sep: Daybreak keys passed, DeepSeek V4.1-Flash live with Pro remap, Astra and Agents API surfaces landed — Gemini 4 and Apple China launch timing still open.
NVIDIA launched DSX MaxLPS, a full-stack platform that dynamically reallocates stranded power across GPU racks to fit up to 40% more compute in the same facility envelope. The system combines real-time power steering, workload-aware profiles, and 45°C liquid cooling to boost performance per watt by 1.3–1.5x on Blackwell and Vera Rubin systems.
LTX, the open world model company spun out of Lightricks, has released LTX-2.5 — a rebuilt video and world model that generates a 10-second, 720p clip. The release reinforces a business model bet that open weights — not closed APIs — will.
Z.ai released GLM-5.2, a 753B open-weight model with MIT licensing that matches Anthropic's Opus 4.8 on agentic benchmarks at roughly one-fifth the token cost. The 1M-context model is already running in Cursor and Claude Code, shifting enterprise leverage toward self-hosted inference.
OpenAI published measured Jalapeño inference results (1.5–1.9× performance per watt on InferenceX) and, in a companion full-stack strategy post, detailed how chips, partners, data centers, and serving software compound to lower the cost of useful intelligence.
SpaceX is making one of the largest hardware investments in AI history. SpaceX's AI ambitions go beyond its own needs.