OpenAI unveils GPT-6.2 Astra and Sol with performance gains and lower API costs
OpenAI launched two new models, GPT-6.2 Astra and GPT-6.2 Sol, delivering better reasoning, coding and scientific results while cutting API fees for developers.
Key takeaways
- GPT-6.2 Astra scores 5.2 points higher than GPT-6 Astra on the Agents’ Last Exam benchmark.
- GPT-6.2 Sol runs at roughly one‑fifth the price of Astra and one‑seventh the price of its predecessor for API usage.
- OpenAI reduced Astra’s API token price by 20% and introduced tighter cybersecurity safeguards.
OpenAI announced the release of two upgraded large‑language models, GPT-6.2 Astra and GPT-6.2 Sol, on 17 November 2026. Both models are positioned to improve professional‑grade reasoning, coding and scientific research while delivering lower processing costs for developers. Astra is marketed for complex, multi‑step assignments, whereas Sol targets everyday professional tasks. The rollout follows OpenAI’s recent partnership with Google Cloud to bring its models to enterprise customers.
Performance benchmarks show measurable gains
According to the company’s release, GPT-6.2 Astra outperformed its predecessor on several internal tests. On the Agents’ Last Exam—a benchmark that evaluates multi‑step professional tasks—the Astra model scored 5.2 percentage points higher than GPT-6 Astra and beat Anthropic’s Claude Fable 5.2 by 5.5 points. In scientific evaluations, Astra posted a nine‑point advantage on the Terminal‑Bench‑Science 0.1 test and set a new high on the Vals Science Index, reflecting stronger capabilities in biological data analysis and workflow research.
Sol delivers everyday efficiency at a fraction of the cost
OpenAI positioned GPT-6.2 Sol as a cost‑effective alternative for routine professional and technical problems. The model performed near‑par with Astra on Cognition’s FrontierCode 2.0 benchmark and achieved a higher score than Astra on the LifeVal evaluation, which measures real‑world task performance. Pricing for Sol remains unchanged from the prior generation—$2 per million input tokens and $10 per million output tokens—while Astra’s API token price was cut by 20%, bringing the input cost to $8 per million tokens and the output cost to $40 per million tokens.
Stricter safeguards and expanded availability
Both models are now accessible through ChatGPT Work, Codex, and the OpenAI API for Plus, Pro, Enterprise and Education users. OpenAI also announced tighter safeguards after the models demonstrated stronger cybersecurity capabilities. Requests that could be used for high‑risk cyber activities are now restricted, though the company will support approved defensive research via its Daybreak program. Enhanced monitoring will detect unauthorized actions during model operation.
What to watch next
Industry observers will monitor how the reduced pricing and performance improvements affect adoption among enterprise developers, especially in sectors such as law, finance and biotech that require high‑accuracy reasoning. Competition is intensifying, with SpaceXAI’s Grok 4.7 and Anthropic’s upcoming frontier models slated for release later this year. Future updates on OpenAI’s safeguard framework and any regulatory scrutiny of advanced AI capabilities will also be key signals for the market.
Sources
Get stories like this in your inbox
The day’s biggest crypto stories and market moves in one short morning email. Free, no spam.