OpenAI has launched GPT-6.2 Astra and Sol and made them available for use.
BIBIBI
AT A GLANCE
According to the news, OpenAI has launched GPT-6.2 Astra and Sol, focusing on professional tasks and daily work respectively, with a price reduction for Astra.
Article
OpenAI launches GPT-6.2 Astra and Sol, improving performance while reducing usage costs
Breaking news: OpenAI has introduced two new models, AI and GPT-6.2Astra, designed for complex professional tasks and daily work, respectively. GPT-6.2Sol enhances capabilities in documents, presentations, spreadsheets, programming, and research, achieving a GPT-6.2Astra improvement over the previous generation in multi-step professional task benchmarks, with execution costs less than one-third; 5.2 input and output prices have both been reduced by API, and cache input prices have been lowered by 20%. Its 60% pricing is per million token inputs, with API rates for inputs, 8 USD, cached inputs, 0.4 USD, and outputs, 40 USD。
GPT-6.2Sol focusing on everyday tasks and multi-step workflows, while API prices remain at the previous generation level—per million token inputs, 2 USD, cached inputs, 0.1 USD, and outputs, 10 USD. Additionally, both models strengthen abilities in computer operations, web design, 3D modeling, and cybersecurity, and are now available to eligible ChatGPTWork and Codex users, offering access via API.
Breaking news reports that OpenAI has launched GPT-6.2 Astra and GPT-6.2 Sol, targeting complex professional tasks and daily work, respectively.
02
According to the report, Astra enhances document, presentation, spreadsheet, programming, and research capabilities; its multi-step professional task benchmark improved by 5.2 points compared to the previous generation, with execution costs less than one-third.
03
According to reports, Astra's API input and output prices have both been reduced by 20%, while the cache input price has been reduced by 60%. The prices for million word elements input, cache input, and output are respectively 8 USD, 0.4 USD, and 40 USD.
04
According to reports, Sol focuses on daily tasks and multi-step workflows, and the API prices remain at the same level as previous generations; the prices for million word tokens, cached input, and output are 2 USD, 0.1 USD, and 10 USD respectively.
05
Reports say that both models enhance computer operations, web design, 3D modeling, and network security skills. They have been made available to eligible ChatGPTWork and Codex users, with API access provided.
AI-assisted interpretation
The following is analysis, separate from reported facts. Verify important claims independently.
The news report positions the two models for different purposes: Astra is designed for complex professional tasks, while Sol is for daily work. The report lists the billing prices for input, cached input, and output separately; Astra has a price reduction, while Sol’s price remains unchanged. Therefore, it cannot be assumed that both models have seen price reductions.
Why it matters to readers
The report also covers task capabilities, usage prices, and scope of use. These can serve as clues for comparing the two models, but confirmation from the original announcement is still needed.
First, determine whether your needs are part of daily work or complex professional tasks. Then, check the input, cached input, and output prices separately. The reported open range is limited to eligible users only.
Risks and unknowns
The provided evidence is from media news, and it does not include OpenAI’s original announcement. The model’s release, capabilities, price, and open scope have not been independently verified.
No name for the professional task evaluation, scoring scale, or testing conditions are specified, so it is impossible to determine the actual significance of achieving a score of 5.2.
Without specifying the measurement criteria and conditions under which “less than one-third of the execution cost” is achieved, it cannot be directly equated with a price reduction for API.
No specific qualifications for ChatGPTWork and Codex users are specified.
Related Developments
Loading event timeline…
Related concepts
API
This term is not in the glossary yet. Browse related concepts in the glossary.