Data from Artificial Analysis reveals that the cost to reach specific 'Intelligence Index' thresholds is dropping rapidly. Reaching an index of 30 now costs 86x less per task than when first achieved, with GPT-6 Luna hitting it at $0.03/task. While top-tier models like Claude Opus 5.5 command higher absolute costs due to expanded capability ceilings, the barrier to entry for mid-level intelligence has significantly decreased.
Termsslug: adaptive-reasoning
Adaptive Reasoning
自适应推理
Mechanism where models dynamically adjust thinking depth based on task complexity
Curated Intelligence
Artificial Analysis (@ArtificialAnlys)T1·Models
78 pts
OpenAI 官方博客T1·Models
68 pts
OpenAI has officially released GPT-5.1 in the API, featuring faster adaptive reasoning and improved coding performance. The update introduces extended prompt caching and native tools like apply_patch and shell to optimize developer workflows.
OpenAI 官方博客T1·Models
45 pts
OpenAI released a system card addendum for GPT-5.1 Instant and Thinking, updating baseline safety metrics. The document primarily supplements evaluation data on sensitive conversations, emotional reliance, and mental health risks, while core capability descriptions remain consistent with previous blog posts.