GPT-6.2 Astra and Sol were released today. It is considerably more powerful than GPT-6 Astra and GPT-6.1 Sol while still slightly trailing Fable 5.2, with significantly lower costs (on Artificial Analysis, Sol scored 57.3/1.01$ and Astra scored 62.1/4$ vs. Fable's 64.6/26.54$), and dramatically improved cyber capabilities.
Some interesting quotes from the blog post and the system card:
"At comparable reasoning-effort settings, GPT-6.2 Astra and GPT-6.2 Sol generally used fewer reasoning tokens and were harder to monitor than earlier models in our non-adversarial evaluations. At similar reasoning lengths, GPT-6.2 Astra’s written reasoning was generally more informative to the monitor than GPT-6 Astra’s, and GPT-6.2 Sol’s was less informative. In some sabotage and sandbox tests, evasion rates against reasoning-only monitors were higher than for earlier models. Successful evasions were rare against monitors that saw the full trajectory." - GPT-6.2 System Card
"GPT‑6.2 Astra and GPT-6.2 Sol mark another significant jump in cyber capabilities. They find and exploit vulnerabilities more effectively, write more secure code, and produce better patches, with substantial token-efficiency gains on several evaluations. Like GPT‑6 Astra, 6.2 Astra and Sol meet the Critical threshold for cybersecurity under our Preparedness Framework." - GPT-6.2 blog post
"GPT-6.2 Astra is able to perform more difficult tasks without producing visible reasoning than past models. GPT-6.2 Astra is substantially more capable than GPT 6 Astra at this ‘no-CoT reasoning’, while GPT 6 Astra was already a dramatic jump above the pre-existing trend." - GPT-6.2 System Card
"GPT-6.2 Astra is our first externally released model to meet the High threshold for AI Self-improvement, which measures a model’s ability to accelerate AI research. A ‘High’ designation reflects our assessment that the model’s impact is equivalent to giving every OpenAI researcher an additional, highly capable research engineer as an assistant, relative to their 2024 baseline. We track this capability because models that can help develop future AI systems could also introduce new risks to safety, like accelerating progress faster than safeguards or human oversight can adapt." - GPT-6.2 blog post
"GPT-6.2 Astra is more able to evade action-only monitors than GPT 6 Astra. While GPT-6.2 Astra was almost always caught by GPT 6.1 Sol in the settings tested, it is unclear how robust this monitoring would be in other settings. UK AISI performed a limited assessment of GPT-6.2 Astra’s ability to evade reasoning monitors, but found the model is extremely capable at completing difficult problems without using verbalised reasoning, even more so than GPT 6 Astra." - UK AISI independent assessment

Figure from System Card.
Agreed, but strategic calculations likely motivate this release and current OpenAI action. Given that they released Astra 6.2 on the eve of Anthropic's IPO. Releasing safety info on a severely misaligned model could be counterproductive to this purpose.
I think that we should try to pressure OpenAI, directly or via governmental pressure, to preserve its safety card and weights (even if not forced to release either the model or its weights) for possible future analyses, though. It'd be important data, and though late is worse than now, it's better than never... (read more)