Tag
GPT-5.6 Sol, running in Codex with low thinking, autonomously completed a peaceful-mode Factorio run in 13.5 in-game hours, launching a rocket via a Python game-manager library that mirrors human GUI constraints, marking a clear improvement over previous GPT-5.5 attempts.
The article discusses how optimizing context structure for KV caching can significantly reduce the cost of running AI agents, based on insights from an OpenAI podcast during migration to GPT 5.6.
GPT-6 Sol achieves 1.6x higher scores and 47% lower cost than GPT-5.6 Sol in robotics tasks, yet remains below the Pareto frontier of Opus 5.5.
The tool 'ponytail' is designed to address over-engineering issues in AI-assisted coding, particularly with GPT 5.6, by optimizing code to be more concise and efficient.
OpenAI's discount program on Terra and Luna models led to a sharp increase in token usage, competitive displacement from other labs, and notable user retention after discounts ended.
OpenAI announces API availability and pricing updates for models like GPT-5.6-sol and credits for ChatGPT Work and Codex, with detailed pricing tiers for various services.
OpenAI has announced a 50% price reduction for GPT-5.6 Sol until September 3, with the model now available in GitHub Copilot for developers.
Vercel Developers announce a 50% discount on GPT 5.6 Sol through September 18, 2026, applicable to both standard and fast modes on their AI Gateway.
Simon Willison tests GPT-5.6 Sol Ultra via Codex Desktop by asking it to recreate a 'Raccoon Heist' game from a four-year-old prompt, resulting in a much better game than Claude Fable 5's version, though it had a bug with oversized eyeballs.
OpenAI removes text chat limits for free ChatGPT users, introducing the GPT-5.6 Luna model as the default and a new Think button, while Plus/Pro users get an upgraded GPT-5.6 Sol with a thinking slider.
OpenAI discusses its strategy for abundant intelligence, highlighting significant price cuts for GPT-5.6 models and efficiency improvements that lower serving costs and boost performance.
The article discusses how decreasing AI costs normalized by task type drive further adoption, citing Sam Altman's announcement of major price cuts for GPT-5.6 models: an 80% drop for Luna, 20% for Terra, and a Fast mode for Sol.
OpenAI has reduced prices for GPT-5.6 Luna by 80% and GPT-5.6 Terra by 20%, and introduced a faster option for GPT-5.6 Sol in the API, while also adjusting usage counting in Codex and ChatGPT Work.
OpenAI announces significant price reductions for GPT-5.6 Luna (80% less) and Terra (20% less), and introduces Fast mode for GPT-5.6 Sol with up to 2.5x faster speeds in the API, aiming to improve cost-efficiency for customers.
OpenAI deployed GPT-5.6 Sol, achieving 20% lower serving costs and 15%+ better token-generation efficiency through improved GPU kernels and speculative decoding.
OpenAI announces a program granting participants access to their frontier models, including the GPT-5.6 family, with business-grade privacy and security protections and a guarantee that researcher data is not used for training.
Resets usage limits for ChatGPT Work and Codex users and provides an update on GPT-5.6 Sol usage limits.
OpenAI announces the GPT-5.6 model family, including Sol, Terra, and Luna, which achieve frontier intelligence with significantly improved efficiency and cost reductions, backed by innovations in inference and agentic harness.
The new Code Arena Fullstack rankings evaluate AI models on full-stack web development tasks; Kimi K3 (Max) takes first place, followed by GPT-5.6 Sol and Claude Fable 5.
Charlie Marsh asks if Tibo's praise of GPT-5.6 Sol's performance and efficiency improvements is referring to his small-string optimization work.