OpenAI's GPT-5.6 Sol matches top models at one-third cost

OpenAI’s new flagship model, GPT-5.6 Sol, is turning heads by matching the performance of Anthropic’s Claude Fable 5 on key benchmarks while costing only a third of the price. On the Intelligence Index, Sol scores 59, just one point behind Fable 5, and it actually beats all other models on the Coding Agent Index with 80 points in OpenAI’s Codex environment. It also leads in “Presentation Elo” for realistic office tasks. The pricing is aggressive: Sol costs $1.04 per task versus Fable 5’s higher rate. Cheaper variants—Terra and Luna—slash costs even further. This is part of a broader trend: Chinese open models, Meta’s Muse 1.1, and xAI’s Grok 4.5 are all forcing price cuts across the industry. OpenAI is still pricier than those competitors, but it’s putting real pressure on Anthropic to drop its prices, too. Why it matters: This pricing war may ultimately squeeze margins across the sector, forcing AI labs to compete on efficiency rather than just raw capability.