OpenAI reports Jalapeño chip delivers up to 1.9 times more inference work per watt
OpenAI published measured results for its Jalapeño inference chip across gpt-oss-120b, DeepSeek-R1, and Kimi K2.5. It reports 1.5 to 1.9 times more work per watt at peak throughput, 1.7 to 3.6 times lower end-to-end latency, and plans to begin deployment in its compute infrastructure by year-end.
0 comments · 0 votes
Sign in to join the discussion →
No comments yet. Start the discussion.
Why it moved the index
The measured efficiency and latency gains make high-volume inference and responsive agents cheaper to scale, strengthening a frontier developer's full-stack position and competitive pressure to deploy more capable systems. The performance measurements and deployment plan are primary evidence. The risk nexus requires one inference from lower serving cost and latency to faster diffusion and competitive acceleration, so confidence is capped at 60.
Assessment history
-
R1
Toward 54 · confidence 60
New measured performance results and a deployment timetable materially extend the earlier Jalapeño announcement.
25 Aug 2026
Share this page
-
DoomBench assesses “OpenAI reports Jalapeño chip delivers up to 1.9 times more inference work per watt” as evidence moving toward doom, with magnitude 54 and confidence 60 out of 100 in the competitive race category.
-
The DoomBench assessment of “OpenAI reports Jalapeño chip delivers up to 1.9 times more inference work per watt” is based on reporting from OpenAI and records the editorial rationale, source quality, attribution, and revision history.
-
DoomBench summarizes “OpenAI reports Jalapeño chip delivers up to 1.9 times more inference work per watt” as follows: OpenAI published measured results for its Jalapeño inference chip across gpt-oss-120b, DeepSeek-R1, and Kimi K2.5. It...
https://www.doombench.com/news/openai-reports-jalapeno-chip-delivers-up-to-1-9-times-more-inference-work-per-watt-2026-08-25