Skip to content
VECTOR WIREAI INTELLIGENCE
UTC
Home

OpenAI cuts Astra API costs 20% with GPT-6.2, matches Anthropic on evals

GPT-6.2 Astra API priced at $8/1M input, $40/1M output — 20% below GPT-6 Astra; cached input 60% lower.

GPT-6.2 Astra is priced at $8 per million input tokens, $0.40 per million cached input tokens, and $40 per million output tokens4. Input and output rates are 20% below GPT-6 Astra's, while cached input pricing is 60% lower. GPT-6.2 Sol holds GPT-6.1 Sol's pricing at $2 per million input tokens and $10 per million output tokens.

Benchmark picture

On Agents' Last Exam, which tests multistep professional workflows, GPT-6.2 Astra scored 5.2 percentage points above GPT-6 Astra at less than one-third the cost, and 5.5 points ahead of Anthropic's Claude Fable 5.2 at 5% of the cost, according to OpenAI10. On Terminal-Bench-Science 0.1, GPT-6.2 Astra scored 9 percentage points higher than GPT-6 Astra. OpenAI said GPT-6.2 Astra also set a new high on the Vals Science Index.

On OpenAI's internal GDPval v2 evaluation, GPT-6.2 Astra improved by 6.2 percentage points while using 22% fewer output tokens, and GPT-6.2 Sol improved by 5.3 points over GPT-6.1 Sol9. OpenAI said GPT-6.2 Astra more than doubled Claude Fable 5.2's pass rate on Harvey LAB-AA legal tasks and ranked first on AA-3D with 1,318 Elo, ahead of Claude Fable 5.2 at 1,271 and GPT-6.2 Sol at 1,259.

Third-party evaluators offered a more mixed view. In the Artificial Analysis Intelligence Index, GPT-6.2 Astra ranked second, trailing only Claude Fable 5.2, but Artificial Analysis said it dominates the Pareto frontier in cost-performance ratio5. Artificial Analysis showed GPT-6.2 Astra flat versus GPT-6 Astra in coding7. Epoch AI placed GPT-6.2 Astra tied with Claude Fable 5.2.

Sol closes the gap on code

GPT-6.2 Sol came within 0.5 points of GPT-6.2 Astra on Cognition's FrontierCode 2.0 coding evaluation at approximately one-fifth the cost when both models used maximum reasoning, according to OpenAI8.

Both models are available to Plus, Pro, Enterprise, and Edu users in ChatGPT Work and Codex, and to developers through the API as gpt-6.2-astra and gpt-6.2-sol. Pro-500 subscribers can use Ultrafast mode with both models, and for the first time, Pro-200 subscribers also get Ultrafast access, starting with GPT-6.2 Sol.

Context and competitive timing

OpenAI kicked off the GPT-6 generation with Astra in September, then upgraded Sol to GPT-6.1 at DevDay. GPT-6.1 Astra was canceled over safety concerns, according to Digital Trends. The release comes after Anthropic launched Claude Fable and Mythos 5.2 and Google released Gemini 4 Hydrogen, according to 9to5Google11.

ANALYSIS The pricing and efficiency story may matter as much as the benchmark numbers. GPT-6.2 Astra's 20% input/output price cut and 60% cached-input reduction, combined with 22% fewer output tokens on GDPval v2, compound into a substantially lower per-task cost for API consumers. Third-party results that show GPT-6.2 Astra flat in coding and trailing Claude Fable 5.2 on the Artificial Analysis index temper OpenAI's own eval claims, leaving the cost-performance ratio as the clearest differentiator.

Was this useful?

The day’s vector, in your inbox.

Vector Wire Daily: one dispatch each publishing day.