← The Polylog AI Intelligence Brief
Morning Edition · Sunday, August 9, 2026Published at 3:02 AM EDT · New York
Claude Opus 5 Posts 30.2% on ARC-AGI-3, Roughly Four Times the Next-Best Published Score
Anthropic held pricing at $5 and $25 per million input and output tokens while reporting 96.0% on SWE-bench Verified and a jump from 1.5% on the abstract reasoning benchmark.
Anthropic's Claude Opus 5, released on July 24, remains the reference point every competitor's benchmark chart is drawn against, including the one Meta published with Muse Code this week. The reported numbers are worth restating precisely.…
Continue the AI Intelligence Brief
Track frontier labs, chips, export controls, model releases, regulation, and AI infrastructure.
- 5 AI intelligence signals a day
- Frontier labs, compute, and chips
- Model releases and AI infrastructure
- Source-grounded analysis with confidence labels
The Global Intelligence Brief stays free.
Part of a tracked trend
Frontier Model Efficiency Gains
Capability per unit of training and inference compute keeps improving, letting newer models match prior frontier performance far more cheaply and gradually loosening the link between raw scale and capability.
Townsquare
Start a discussion in Townsquare.
More from this edition
- OpenAI Slows Astra Release After Internal Tests Point to Critical Cyber Capability
- SpaceX and Tesla Commit $16.8 Billion to a Texas Chip Plant Built for Their Own Silicon
- Amazon Is the Buyer Behind a 7.65-Gigawatt Texas Gas Plant Permitted for 33 Million Tons of Carbon Dioxide
- Meta's Muse Code Claims 82.9% on Terminal-Bench, a Number the Verified Leaderboard Does Not Show
- Armenia Opens a 15-Megawatt NVIDIA Cluster That Exists Because Washington Granted an Export Licence
- Two Open Meta Vision Models Cut Beamline Data Analysis From a Month to Fifteen Minutes at Berkeley Lab
- A $41.5 Million Robotic Wheelchair Program Runs Meta's Vision Models Directly on Battery-Powered Hardware
Comments
0No comments yet.