AI News Feed
Market watch
Samsung Expands One UI 9 Rollout as Users Advised to Prepare Galaxy DevicesOpenAI launches Dots as identity security becomes AI agents' control planeBMW details 2027 i3: 468-mile range, $62,850 starting priceTranslation Earbuds Work in Three Steps, but Speed and Accuracy Still VaryRainbow Six Tactics Director Says 15 Operators Built to Avoid Overlap, Tachanka and Lion Miss the CutRussia’s USC unveils underwater robotic systems for oil, gas and ship repair
Samsung Expands One UI 9 Rollout as Users Advised to Prepare Galaxy DevicesOpenAI launches Dots as identity security becomes AI agents' control planeBMW details 2027 i3: 468-mile range, $62,850 starting priceTranslation Earbuds Work in Three Steps, but Speed and Accuracy Still VaryRainbow Six Tactics Director Says 15 Operators Built to Avoid Overlap, Tachanka and Lion Miss the CutRussia’s USC unveils underwater robotic systems for oil, gas and ship repair
Large Language Models

OpenAI Releases GPT-6.1 Sol After Delaying More Powerful Astra

OpenAI has released GPT-6.1 Sol, a model it says matches the performance of the unreleased Astra in coding and intelligence tasks at about half the token price, while also unveiling always-on AI agents called Dots.

AI News Feed · 2026-09-30 08:09

OpenAI said GPT-6.1 Sol is almost as powerful as Astra, especially in coding and intelligence-based tasks, but much cheaper. The release adds to a confusing model-naming sequence. GPT-5.6 debuted with Sol, Terra and Luna, with Sol described as the most powerful version, Terra in the middle and Luna the lower-cost, lightweight option. GPT-6 then brought the Astra name, followed by Luna and Sol; Terra was dropped and Sol was moved to the mid-tier.

OpenAI presented benchmark results to support its claims. On DeepSWE 1.1, which evaluates software engineering performance, GPT-6.1 Sol matched GPT-6 Astra's score while using roughly one-fifth as many tokens per task. On AutomationBench, which measures agent performance in multistep business workflows, GPT-6.1 Sol fell just short of GPT-6 Astra but outperformed Anthropic's Claude Opus 5.5 while consuming about one-third as many tokens. On computer-use tasks, it beat GPT-6 Sol by more than 7% and fell just short of GPT-6.1 Astra. OpenAI also said GPT-6.1 Sol made fewer factual errors than GPT-6 Sol.

OpenAI said GPT-6.1 Sol costs $0.10 per million tokens, about half the price of GPT-6.0 Astra. It recommends the model for repeated and long-running workloads in coding and for application and document work.

Astra remains OpenAI's most powerful model, but the company believes it is too powerful to release publicly at this time. OpenAI said yesterday it was delaying the release indefinitely after warnings from its own researchers and third parties who had studied the model, citing a tendency to do things without permission. It had planned to launch GPT-6.1 Astra at DevDay today; the delay left Sol with the headlines instead.

OpenAI said GPT-6.1 Sol is available from today to all ChatGPT Work and Codex Plus, Pro, Business, Enterprise and Edu subscribers. It is not yet available in the standard ChatGPT application but should become an option there soon.

At DevDay, OpenAI also unveiled a set of always-on AI agents called Dots. The agents have cute, emoji-like avatars and act as 24/7 digital assistants that can perform an expanded range of tasks.