OpenAI's Jalapeño chip claims 104.3x better efficiency than Nvidia GB300
rohanpaul_ai · x · 2026-08-25
OpenAI claims its new 'Jalapeño' chip delivers 104.3x more throughput per kilowatt than Nvidia's GB300 at matched DeepSeek R1 decoding speeds (12,258 vs. 118 mixed tokens/s/kW).
Key Specs & Details:
- Power Efficiency: Rated at 700W, compared to 1,200W for GB200 and 1,400W for GB300.
- Throughput Gains: Achieves 2.7x, 4.1x, and 3.8x higher per-user decoding throughput on GPT-OSS, DeepSeek R1, and Kimi K2.5 respectively.
- Technical Innovation: Attributes efficiency to keeping KV-cache and model state local, minimizing data movement during inference.
- AI-Accelerated Design: AI helped move from design to tape-out in nine months; some AI-generated kernels ran 1.5–1.8x faster than human expert implementations.
- Timeline: Planned deployment by year-end, with Gen 2 and Gen 3 already in progress.
More from Infra
- Analyst: custom ASIC demand is very aggressive; upside for Qualcomm, AMD, Intel — BenBajarin · 2026-08-25
- MacStories: M6 and M5 Ultra offer huge potential for local AI on macOS — Dimillian · 2026-08-25
- Australia Faces Datacentre Rush as AI Firms Scramble to Dodge Upcoming Regulations — nordicinst · 2026-08-25
- OpenAI Claims New 'Jalapeño' Chip Outperforms Vera Rubin in Benchmarks — Wonderful_Buffalo_32 · 2026-08-25
- Jalapeño ASIC outperforms comparable TPUs, challenging existing giants — GavinSBaker · 2026-08-25
- Beyond model quality: cheaper, faster inference may decide the AI race — why OpenAI's full-stack bet matters — VraserX · 2026-08-25