OpenAI's GPT-6.1 Sol nearly matches Astra at 80% lower cost
What's the story
OpenAI has launched its latest artificial intelligence (AI) model, the GPT-6.1 Sol, at OpenAI's DevDay event on Tuesday. The new model is said to deliver nearly the same level of intelligence as GPT-6 Astra, but at a fraction of the cost. However, unlike what was initially expected, OpenAI has not launched GPT-6.1 Astra due to safety concerns raised during internal testing.
Enhanced performance
GPT-6.1 Sol offers significant improvements over GPT-6 Sol
The new model, GPT-6.1 Sol, is said to offer major improvements over its predecessor in complex tasks like programming and debugging, document understanding, and executing multi-step workflows.
OpenAI claims that on several fronts, this model comes close to the performance of GPT-6 Astra while costing only one-fifth of the standard input/output token prices.
Enhanced accuracy
Error rate for factual inaccuracies has dropped significantly
OpenAI claims that GPT-6.1 Sol shows significant improvements over GPT-6 Sol, especially when it comes to factual accuracy with difficult prompts.
The error rate for factual inaccuracies at low reasoning effort has dropped from 11.4% to 7.7%.
Across all reasoning settings, the new model's error rate remains within 1.9% of that of GPT-6 Astra, according to OpenAI data.
Trustworthiness
The new model is more reliable in respecting user intent
The new model, GPT-6.1 Sol, is said to be more transparent about its limitations and more reliable in respecting user intent and safety constraints.
In difficult tests, it fails less often than its predecessor at flagging broken search tools, following explicit restrictions, and avoiding unauthorized outcomes during tasks.
OpenAI claims it observed no attempts to circumvent the automated safety reviewer, consistent with GPT-6 Astra and GPT-6 Sol.
User access
Availability and commitment to safety
OpenAI has made the new model available from today for all Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex. However, it's worth noting that the model is not yet available in Chat.
The company continues to emphasize its commitment to safety and user trust with these AI advancements.