GPT-6 Astra Scores 67 on Coding Index, Trails Fable 5.1
GPT-6 Astra's strong token efficiency places it on the Pareto frontier for coding cost versus score, even as higher API prices weaken its standing on the Intelligence Index.
Reporting from 1 source: GIGAZINE.
Artificial Analysis benchmark results for OpenAI's GPT-6 Astra show a 67-point score on the Coding Agent Index, tying Fable 5 and trailing Fable 5.1's 70 points. GPT-6 Astra uses about one-third the tokens of GPT-5.6 Sol at maximum reasoning, lowering cost per task below Fable 5. On the Intelligence Index, it scores 61 points, matching GPT-5.6 Sol and Muse Spark 1.3, with cost efficiency hurt by higher API pricing.
Artificial Analysis published GPT-6 Astra results on September 3, 2026, covering both the Coding Agent Index and the Intelligence Index. The coding score of 67 points ties Fable 5 and comes close to Claude Opus 5 at 68 points, while Fable 5.1 leads with 70 points.
Token efficiency is the main differentiator. At maximum reasoning on Codex, GPT-6 Astra consumes roughly one-third the tokens of GPT-5.6 Sol and about one-seventh of Claude Opus 5. That efficiency cuts cost per task to less than half of Fable 5's, despite both scoring 67 points. GPT-6 Astra therefore sits on the Pareto frontier, meaning it offers the highest score for its cost or the lowest cost for its score among tested models.
The Intelligence Index tells a different story. GPT-6 Astra's 61 points tie GPT-5.6 Sol and Muse Spark 1.3, five points behind Fable 5.1. Its API pricing of $10 per million input tokens and $50 per million output tokens is 2.5 times GPT-5.6 Sol's, raising cost per task by 75 percent.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.