Abstract:
OpenAI officially released two new artificial intelligence models, GPT-6.2 Astra and GPT-6.2 Sol, and claimed that the new models have achieved significant progress in multiple fields such as programming, scientific research, professional office, and computer operations. At the same time, the new model can complete more tasks with less token consumption, allowing users to achieve higher work efficiency within the existing usage quota.

In September this year, OpenAI launched GPT-6 Astra and called it a generational leap in capabilities. However, as competition intensifies, other AI labs are quickly closing the gap. Among them, Anthropic's Claude Fable 5.2 surpassed Astra in multiple programming benchmarks, while Google's Gemini 4 Carbon provided highly competitive performance at a lower price.
Faced with market challenges, OpenAI launched GPT-6.2 Astra this time, with the goal of further improving cost-effectiveness while regaining its leadership position in performance.
According to data released by OpenAI, in the Agents' Last Exam test that evaluates the ability of complex multi-step professional workflows, GPT-6.2 Astra's performance improved by 5.2% compared to GPT-6 Astra, while the execution cost was less than one-third of the latter. Compared with Claude Fable 5.2, GPT-6.2 Astra's performance is 5.5 percentage points higher, while the cost of a single task is only about 5% of the other party's. In the Terminal-Bench-Science 0.1 scientific test, the new model also scored 9% higher than GPT-6 Astra, while the cost was still less than one-third of the latter.
In addition to the improvement in benchmark test scores, OpenAI also stated that the practicality of GPT-6.2 Astra has been further enhanced in professional office scenarios. The model can produce PowerPoint presentations, Excel spreadsheets, and various documents with higher design quality and more accurately adhere to existing templates and industry specifications. At the same time, its computer operation capabilities and visual reasoning capabilities have also been upgraded.
For the field of software development, OpenAI claims that GPT-6.2 Astra can write clearer code structure, make more reasonable software architecture decisions, and provide more targeted testing solutions, thereby reducing the number of iterations required to complete complex development tasks.
Unlike GPT-6.2 Astra, which is mainly aimed at complex professional work, GPT-6.2 Sol is aimed at daily office and general productivity scenarios. On the FrontierCode 2.0 programming benchmark, GPT-6.2 Sol trailed GPT-6.2 Astra by only 0.5 percentage points, but the per-task cost was only about one-fifth of the latter. In the LifeVal benchmark developed by OpenAI, Sol even outperformed GPT-6 Astra, while costing only about one-seventh of the latter.
In terms of pricing, OpenAI announced that the input and output token prices of the GPT-6.2 Astra standard API will be reduced by 20% compared to GPT-6 Astra, and the cache input token price will be reduced by 60%. GPT-6.2 Sol maintains the same price level as GPT-6.1 Sol.
OpenAI stated that the two new models are available to ChatGPT Plus, Pro, Enterprise and Edu subscribers from now on, and can be used through ChatGPT Work and Codex. Among them, Pro 500 subscribers can enable Ultrafast super-high-speed mode on both models, while Pro 200 subscribers can now use this feature on GPT-6.2 Sol.
As the competitive pressure brought by Anthropic Claude Fable 5.2 and Google Gemini 4 Carbon continues to increase, the launch of GPT-6.2 Astra and GPT-6.2 Sol is regarded as a new round of competitive offensive launched by OpenAI in terms of performance, efficiency and cost, striving to re-establish its leading position in the high-end artificial intelligence model market.
Comments