GPT-6 Astra matches Fable 5 on coding at half the cost, but priced 2.5x its predecessor
Polymarket · x · 2026-09-04
Artificial Analysis published first benchmarks for GPT-6 Astra. Pricing: $10/$50 per million input/output tokens — 2.5x GPT-5.6 Sol's $4/$20, with the same 90% cache-read discount and 25% cache-write premium.
Coding Agent Index
- Scores 67 in Codex, roughly matching Claude Opus 5 and Fable 5; leader Fable 5.1 scores 70
- Major token-efficiency gains: uses 1/3 the tokens of GPT-5.6 Sol (max) and 1/5 of Claude Opus 5 (xhigh)
- Less than half the per-task cost of Claude Fable 5 at equal score
Intelligence Index
- Scores 61, equal to GPT-5.6 Sol, 5 points below Claude Fable 5.1, trailing Meta's Muse Spark 1.3 (max)
- 10% fewer output tokens at max effort, but the 2.5x price makes it 75% more expensive per task
- Hallucination rate drops from 92% to 51% while accuracy rises 4 points (AA-Omniscience)
- 80-point Elo gain on AA-Briefcase, a multi-week long-horizon knowledge work eval
Net: strong coding cost-efficiency, but general intelligence comes at a steep premium.
Related event: GPT-6 Astra Reviews: Halved Hallucinations but 2.5x Price Hike(18 posts)→
More from Models
- OpenAI launches GPT-6 Astra, claiming it can do anything you do on a computer — kagigz · 2026-09-04
- Databricks evals: GPT-6 Astra claims SOTA on OfficeQA Pro benchmarks, cheaper per task — downingARK · 2026-09-04
- Mathematician tests GPT-6 Astra: live Lean proof verification while writing arguments — teortaxesTex · 2026-09-04
- Tavus Launches Sparrow-2, Claiming #1 in End-of-Turn Detection and Interruption Handling — ycombinator · 2026-09-04
- ARC-AGI-3: 10x reasoning tokens cuts total cost from $48k to $26k vs medium — i_dg23 · 2026-09-04
- Researchers flag data contamination concerns in benchmark behind Astra's time-horizon score — dfrsrchtwts · 2026-09-04