OpenAI released GPT-5.6 in three tiers, Sol, Terra, and Luna, with Sol priced at $5/$30 per million input/output tokens and claiming an 80-point score on the Artificial Analysis Coding Agent Index, 2.8 points above Anthropic's Fable 5 while using less than half the output tokens and costing roughly one-third less. CEO Sam Altman cited 54% token efficiency improvement on coding tasks versus prior versions, and the company positioned cybersecurity capability as a headline feature after a reported government-requested rollout delay. Structurally, this release intensifies the inference cost and capability benchmarking competition between OpenAI and Anthropic at the enterprise layer, with pricing and efficiency claims directly relevant to whether app-layer or model-layer players capture margin. Also reported by: https://www.cnbc.com/2026/07/08/openai-expanding-gpt-5point6-ai-model-release-ending-government-limits.html
587 other tracked developments touching the same claims.