OpenAI says GPT‑6.1 Sol is its most demanded model ever

OpenAI engineer Tibo Sottiaux said on X that GPT-6.1 Sol is "our most demanded model pretty much ever" across the API and subscriptions. ChatGPT and Codex ran under heavy load, he said, and OpenAI has brought more capacity online. He said speed should improve over the coming hours, reaching almost twice what it served the day before.

Latent Space, which covered DevDay, reports the model is priced at $2 per million input tokens and $10 per million output, with cached input at $0.10. It describes OpenAI's pitch as near-Astra intelligence for a fifth of the price, and says OpenAI claims a tie with GPT-6 Astra on DeepSWE. OpenAI also says it beats Claude Opus 5.5 on AutomationBench at a third of the cost.

Independent numbers point the same way, though not as far. Latent Space cites Artificial Analysis as placing Sol about one point below Astra on its Intelligence Index, at $0.72 per task against $3.26. Artificial Analysis separately said Sol's cost per task is roughly 30% below GPT-6 Sol's, helped by fewer turns and a lower cache-read price.

DevDay carried more than the model. Latent Space lists Dots, always-on voice agents it says run on GPT-6 Astra and connect to more than 4,000 apps, for Pro, Business Premium and Enterprise plans. It also lists an Ultrafast mode that OpenAI says reaches 300 tokens per second in Codex, at six times standard prices, and a ChatGPT figure of 1.2 billion weekly users.

TechCrunch reported that a Decisions API, in limited preview, lets the small GPT-6 Luna model pick from a fixed set of options, which Sam Altman said makes it "extremely fast". Nobody outside OpenAI has published a test of the Dots agents or the Ultrafast speeds, and the demand figure rests on Sottiaux's post alone.