AI Signal 142
GPT-6 Astra equals Fable 5 in coding at half the cost, but 2.5x pricing makes intelligence tasks 75% more expensive
↗
GPT-6 Astra matches top coding models at less than half the per-task cost of Fable 5, but its 2.5x price increase over GPT-5.6 Sol makes general intelligence work 75% more expensive despite token efficiency gains.
For coding-focused workflows, GPT-6 Astra delivers top-tier performance at a compelling price point against competitors. For general intelligence tasks, the price hike overwhelms efficiency gains, making the predecessor a better value. The halved hallucination rate is a meaningful reliability improvement, but several benchmark regressions complicate adoption decisions.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
GPT-6 Astra scores 67 in the Coding Agent Index, equal to Fable 5 and Claude Opus 5, at less than half the per-task cost of Fable 5.
Pricing rose 2.5x over GPT-5.6 Sol (from $4/$20 to $10/$50 per million input/output tokens), making intelligence tasks 75% more expensive despite ~10% fewer output tokens.
Hallucination rate dropped from 92% to 51% at max effort with a 4-point accuracy gain, but GDPval-AA v2 regressed ~80 Elo and other benchmarks saw 2-3 point declines.
THE CLUSTER