GPT-6 Astra on ARC-AGI-3: 62.7% measured by the ARC Prize, far from the 99.9% claimed
Using the standard protocol, the organization that runs the test gets a score well below OpenAI's, at a cost of more than $26,000.
Section
Models & labs
Large models, AI labs, benchmarks and research announcements.
Kimi K3.1 Pro enters the Intelligence Index at No. 4, at $2 per million tokens
Scored at 55 by Artificial Analysis, the model combines a top-tier ranking with aggressive pricing. Its weights are promised by November 4.
Gemini 4 Argon is available to everyone; a “Carbon” version remains a rumor
Announced at the end of September, Google's new model has been open to everyone since October 13. Another variant is reportedly being tested internally, without official confirmation.