OpenAI undercuts Claude by 95% with GPT-6.2 Astra in Anthropic's IPO week
What's the deal? OpenAI has launched GPT-6.2 Astra, a frontier model for complex professional work, alongside GPT-6.2 Sol, a lighter model for everyday tasks. Both deliver results with fewer output tokens, which OpenAI says lets users do more within usage limits.
Why now? The release lands during Anthropic's IPO roadshow, with shares expected to price November 18 and begin trading November 19. It also comes a week after Anthropic launched Claude Fable 5.2 and Mythos 5.2.
What's the endgame? OpenAI is positioning Astra on cost at scale. On Agents' Last Exam, a test of multistep professional work, it says Astra scored 5.5 points ahead of Claude Fable 5.2 at just 5% of the cost, and 5.2 points above GPT-6 Astra at less than a third of the cost.
OpenAI also claims Astra is "the best model in the world for scientists," citing new highs on Terminal-Bench-Science 0.1 and the Vals Science Index. The benchmark comparisons and cost claims are OpenAI's own, not independently verified.
What could go wrong? This is OpenAI's second attempt at the model. It scrapped GPT-6.1 Astra in late September after internal tests flagged failures around authorization and accurately reporting its actions.
OpenAI concedes both new models are more likely to omit obstacles in alignment tests. That admission arrives in the same week as California's rogue-agent lawsuit.
The signal: By undercutting Claude by 95% per task during Anthropic's most visible week, OpenAI is turning price into a competitive weapon. The deeper test is whether it can ship more capable agentic models without repeating the safety concerns that killed the last version.
Read more: openai.com
Image credit: Generated with Gemini