GPT-6 Astra Reviews: Halved Hallucinations but 2.5x Price Hike

On September 4, Artificial Analysis published its first evaluation of GPT-6 Astra, covering pricing, individual components of the Intelligence Index v4.1.1, and cost efficiency. The overall verdict: performance and efficiency improved significantly, but the 2.5x price increase leaves value for money behind the previous generation, with results varying widely across benchmarks.

Confirmed

• Pricing: Input/output rose from $4/$20 for GPT-5.6 Sol to $10/$50 per million tokens—2.5x across the board; the 10% cache read discount and 25% cache write surcharge carry over.
• Hallucination rate: At max effort, the hallucination rate fell from 92% to 51%, the main driver of the big jump in AA-Omniscience.
• Mixed results: AA-Briefcase improved by about 80 points (rubric scores and analysis quality), but GDPval regressed.
• Token efficiency: Versus GPT-5.6 Sol, output tokens per task at max effort dropped roughly 10%, defining a new Pareto frontier on "intelligence index vs. output tokens per task."
• Coding: Tied with Fable 5 on the Coding Agent Index but at lower cost; max-effort token usage is about 3x lower than GPT-5.6 Sol(max), the main source of its cost improvement.
• Cost accounting: Despite lower token usage, cost per task at max effort is still 75% higher than GPT-5.6 Sol.

Why it matters

GPT-6 Astra sets new Pareto frontiers in both intelligence and token efficiency, and the halved hallucination rate is especially noteworthy, signaling real reliability progress. But the 2.5x price increase eats up most of the efficiency gains, leaving overall value for money below the previous generation—whether users pay for the quality boost will depend on the use case (the coding Agent cost advantage is relatively clear).

2026-09-04 ~ 2026-09-04 · 18 related posts

• GPT-6 Astra matches Fable 5 on coding at half the cost, but priced 2.5x its predecessor — Polymarket · 2026-09-04
• GPT-6 Astra Priced at $10/$50 per Million Tokens, 2.5x GPT-5.6 Sol — ArtificialAnlys · 2026-09-04
• GPT-6 Astra Cuts Coding Agent Token Use ~3x at Max Effort vs GPT-5.6 Sol — ArtificialAnlys · 2026-09-04
• GPT-6 Astra Sets New Pareto Frontier, Cutting Output Tokens ~10% vs GPT-5.6 Sol — ArtificialAnlys · 2026-09-04
• GPT-6 Astra gains ~80 points on AA-Briefcase but regresses similarly on GDPval-AA v2 — ArtificialAnlys · 2026-09-04
• [source] GPT-6 Astra cuts hallucination rate from 92% to 51%, driving big AA-Omniscience jump — ArtificialAnlys · 2026-09-04
• Artificial Analysis adds GPT-6 Astra comparison against leading models — ArtificialAnlys · 2026-09-04
• GPT-6 Astra's AA Intelligence Index and Coding Agent Index Scores Surface — signed7 · 2026-09-04
• GPT-6 Astra Early Test: 21B Tokens In, Priced at 2x GPT 5.6 Sol — charliermarsh · 2026-09-04
• GPT-6's Artificial Analysis Intelligence Index scores surface in new gallery — ColorLaser · 2026-09-04
• GPT-6 Astra loses to fable 5.1 and opus 5 on key benchmarks, priced at $50 per 1M output tokens — ns123abc · 2026-09-04
• GPT-6 Astra barely improves over 5.6 Sol on Artificial Analysis Intelligence Index — burny_tech · 2026-09-04
• GPT-6 Astra scores 61 on AA Intelligence Index — no IQ gain over predecessor, price up 2.5x — MicahBerkley · 2026-09-04

5 near-duplicate retellings: ArtificialAnlys · ArtificialAnlys · ArtificialAnlys · cedric_chee · cedric_chee