Source: the-decoder.com/openai-launches-gpt-6-2-astra-and-sol-with-lower-api-prices-and-some-benchmark-gains
Astra is available to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and developers can reach both models through the OpenAI API. Astra costs $8 per million input tokens, $0.40 per million cached input tokens and $40 per million output tokens. Sol keeps the pricing of GPT-6.1 Sol at $2 input, $0.10 cached input and $10 output per million tokens.
OpenAI reports gains on its own benchmarks, including Agents' Last Exam, where Astra scores 5.2 points above GPT-6 Astra at less than a third of the cost. Artificial Analysis ranks Astra second on its Intelligence Index with 62 points, behind Claude Fable 5.2 at 65. Epoch AI finds Astra tied with Claude Fable 5.2 on its capabilities index but weaker on real-world research tasks.
Both models reached the "Critical" cybersecurity threshold in OpenAI's Preparedness Framework, and OpenAI is adding stricter safeguards that will also limit legitimate security testing. OpenAI also says Astra is the first externally released model to reach the "High" threshold for AI self-improvement. Artificial Analysis reports hallucination rates of 59 percent for Astra and 55 percent for Sol on a hard factual benchmark at maximum effort.
Why it matters if you build agents
Developers building agents can now pick Astra for heavier professional workloads and Sol for cheaper, faster tasks, with Vals AI reporting Sol at $3.18 per task against Astra's $10.68 on its index.
Hallucination rates on the Artificial Analysis test mean agents that answer factual questions should verify outputs rather than trust them. Security tooling built on these models may run into stricter refusals for vulnerability research.
Get stories like this every weekday
One short email with the day's agent tooling news.