OpenAI has introduced GPT-6.2 Astra and GPT-6.2 Sol, with Astra’s API input and output token prices 20% below GPT-6 Astra’s and Sol retaining GPT-6.1 Sol’s rates. The company positions Astra for difficult professional projects and Sol for everyday tasks, saying both need fewer output tokens to handle demanding work.
For GPT-6.2 Astra, developers pay $8 per million input tokens and $40 per million output tokens, while cached input costs $0.40 per million tokens. OpenAI puts the cached-input price reduction at 60%. The cheaper Sol model finished within 0.5 points of Astra on Cognition’s FrontierCode 2.0 at roughly one-fifth of its cost, with both models using maximum effort, according to OpenAI.
Read more: OpenAI Plans to Cut Token Prices Before Anthropic Makes the Same Move
FrontierCode 2.0 assesses whether edits to actual codebases satisfy the standards maintainers use when accepting changes. On Agents’ Last Exam, which evaluates extended professional workflows, OpenAI reports that Astra beat GPT-6 Astra by 5.2 points while costing less than a third as much. On its internal GDPval v2 assessment of knowledge work, Astra improved by 6.2 percentage points over GPT-6 Astra and generated 22% fewer output tokens.
OpenAI says Astra produces more direct writing and follows professional conventions more closely when making documents, spreadsheets and presentations. For slides, it says the model can reproduce a supplied template’s typography, colors and layout from either a file or a screenshot. On Artificial Analysis’s AA-AnalystAgent, which uses spreadsheets and reports, Astra answered 62.5% of questions correctly across all five independent attempts, according to OpenAI.
Sol improved by 5.3 points over GPT-6.1 Sol on GDPval v2, OpenAI says. On LifeVal, its evaluation of difficult everyday tasks tailored to individual needs, the company says Sol surpassed GPT-6 Astra at about one-seventh of the cost. Those tasks include comparing homes against a family’s budget and commute, and arranging meals for a parent adapting to a diagnosis.
The releases follow GPT-6.1 Terra, which offered standard API pricing 50% below GPT-5.6 Terra and GPT-6.1 Sol for visual reasoning, writing and computer tasks. At that launch, OpenAI continued to recommend GPT-6.1 Sol and GPT-6 Astra for its most demanding programming uses.













