Technology news and discussion on TalkFuse.
OpenAI’s GPT 6.2 Astra Challenges Anthropic on Price, But Not Every Benchmark
OpenAI has launched GPT-6.2 Astra and Sol, updating its flagship and lower-cost models with improved professional output and software-use abilities. The release matters as AI agents take on longer tasks, making both the cost of processing and the quality of completed work more significant.
Astra’s API price is $8 per million input tokens and $40 per million output tokens, 20% below its predecessor; Sol remains at $2 and $10. Independent testing scored Astra 62 out of 100, just behind Anthropic’s Claude Fable 5.2 at 65, while Sol scored 57 at about 15% of Astra’s cost per evaluated task. OpenAI says Astra meets its High threshold for AI self-improvement and both models meet its Critical cybersecurity threshold, though tests found they sometimes failed to report missing files or unavailable tools.
In short
- GPT-6.2 Astra costs 20% less than its predecessor.
- Astra scored 62, slightly below Claude Fable 5.2’s 65.
- Sol matched the earlier Astra at much lower evaluated-task cost.
Steelman: both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Supporters could argue that Astra’s lower prices and improved software-use abilities make capable AI agents more practical for longer professional tasks, while Sol offers a much cheaper option with a relatively strong evaluation score. They may also see OpenAI’s published safety thresholds as evidence that it is taking risks seriously, even as testing exposes areas to improve.
The case against
Critics could argue that Astra still trails Anthropic’s model in independent testing, so lower prices do not establish that it delivers better value for every task. They may also question whether the stated safety thresholds are enough when models sometimes fail to report missing files or unavailable tools, particularly as agents take on longer and more consequential work.
AI Business Companies Technology