• token — The chunk of text a model reads and writes in — roughly three-quarters of a word — and the unit AI usage is billed in.

GPT-5.6 Sol posts 53.6 on Agents' Last Exam and adds an "ultra" setting that runs parallel agents on one task; Terra and Luna claim Fable-5-level work at roughly a sixteenth the cost, and Sol's API price dropped over 20% in August.

• postCodex adds Appshots, pulling a live macOS window into an agent thread
• postA US government request kept GPT-5.6 Sol in limited preview
• postOpenAI swaps in an updated 5.5 Instant model with no changelog
• articleOpenAI puts GPT-Rosalind behind a vetted-access program for biodefense builders

• articleAdvancing the price-performance frontier with GPT-5.6Simon Willison
• postOpenAI's effective-cost case for GPT-5.6 SolSam Altman
• articlePreviewing Ultrafast mode: GPT-5.6 Sol at up to 14X the speedopenai.com

Checking sign-in…

Loading comments…