GPT-6 Astra Lands at 67 on Coding Index, Behind Fable 5.1's 70

Artificial Analysis clocked GPT-6 Astra at 67 on its Coding Agent Index run through Codex, roughly equal to Claude Opus 5 and Fable 5, and three points behind Claude Fable 5.1 in Claude Code at 70. The benchmark house frames Astra as a new efficiency floor rather than a capability leader: "GPT-6 Astra defines a new Pareto frontier for Coding Agent Index vs Cost per Task," the analysis states.

Per-task economics carry the story. "Per task, the model is less than half the cost of Claude Fable 5, for the same score," Artificial Analysis writes, and it calls Astra "70% more token efficient than GPT-5.6 Sol." Headline pricing still rose to $10/$50 per million input/output tokens, or "2.5x GPT-5.6 Sol's current prices across the board."

On general intelligence, the picture is flatter. GPT-6 Astra scores 61 on the Intelligence Index, tied with GPT-5.6 Sol and five points behind Fable 5.1 at max with fallback. The hallucination rate at max effort dropped from 92% to 51%, though Artificial Analysis also reports "2-3 point regressions on other evaluations…including reductions in τ³-Banking (customer support), SciCode (Python problems in a scientific domain), and AA-LCR (long context reasoning over large documents)."

The writeup landed on September 3, the same day ChatGPT, Claude and Grok all suffered a simultaneous outage. It is our 354th OpenAI story in the last 90 days, part of a dense run of coding-tools coverage this week.