OpenAI Releases GPT-6.2 Astra and Sol Models, Drawing Sharp Criticism from Developers
OpenAI launched GPT-6.2 Astra and Sol models, facing harsh criticism from experienced users regarding usability and regression despite benchmark improvements.

OpenAI has introduced two new artificial intelligence models, the flagship GPT-6.2 Astra and the more affordable GPT-6.2 Sol 12. Released just a week after the arrival of competing models, the proprietary models are available in ChatGPT Work and Codex for subscribers of Plus, Pro, Enterprise, and Edu tiers, as well as via API 12.
Pricing and Specifications
The flagship GPT-6.2 Astra is designed for complex codebases, web applications, and corporate templates, priced at $8 per 1 million input tokens and $40 for output tokens, marking a 20% reduction in input pricing compared to its predecessor 12. GPT-6.2 Sol targets routine programming and multi-step workflows, maintaining its pricing at $2 for input and $10 for output per 1 million tokens 1. Both models feature a context window of 1.05 million tokens with output generation up to 128,000 tokens 1. OpenAI also introduced a high-speed Ultrafast tier 12.
Industry Backlash and Benchmark Performance
In internal evaluations, GPT-6.2 Astra outperformed its predecessor across all nine categories, while Sol led in eight 1. In the Artificial Analysis Intelligence Index, Astra scored 62 points to secure second place overall 1. However, independent developer testing revealed significant shortcomings 1.
Investor and former HyperWrite AI CEO Matt Schumer dismissed the models as "factually unusable" 1. Developer Theo Braun called 6.2 Astra "the worst model OpenAI has ever released," citing issues with code optimization and template repetition 1. Wharton professor Ethan Mollick described the release as OpenAI's first flagship since GPT-4.5 that failed to feel like a clear forward step, noting that the model was "very uneven and strange" 1.
The models were launched a week after competing releases such as Claude Fable 5.2 and Gemini 4 Hydrogen.
Developers and enterprise users will continue to evaluate the stability and practical utility of GPT-6.2 models in production environments.
Written with AI from the cited sources and checked against them. Open the sources to verify.