>_TheQuery
← All Comparisons

Claude Fable 5.1 vs GPT-6 Astra

GPT-6 AstraClaude Fable 5.1
Pricing
Input / output per 1M$10 / $50$10 / $50
Cached input per 1Mabout $1$0.25
Cache writes per 1Mabout $12.50$12.50 / $20
vs predecessor2.5x over GPT-5.6 SolSame as Fable 5
Coding
DeepSWE v1.174.1%67.4%
Terminal-Bench 4.057.9%55.8%
CursorBench 3.2.0Not reported73.4%
AA Coding Agent Index6770
Math and reasoning
FrontierMath Tier 497.6%87.8%
ARC-AGI-399.9%Not reported
Knowledge
GPQA Diamond96.0%93.7%
HLE, no toolsNot reported60.9%
HLE, with toolsNot reported65.0%
Agentic work
OSWorld72.6%*77.9% / 41.7%*
AutomationBenchNot reported31.4%
Context and controls
Context windowabout 1.05M tokens1M tokens
Max output128K tokens128K tokens
Knowledge cutoffApril 2026June 2026
Effort dialLow - MaxLow - High
Safety and access
Restricted programDaybreakTrusted access (Mythos 5.1)
Public-model limitNo exploit PoCs on standard APICyber tasks rerouted from Fable 5.1

Scores are vendor-reported launch-table figures from September 2026 unless marked independent (AA). OSWorld rows compare different task setups. Some OpenAI-chart Claude rows use reduced-safeguard Mythos rather than shipped Fable 5.1. Prices are list rates - task cost depends on effort level, caching, and harness.

Sources

  1. Introducing Claude Fable 5.1 and Claude Mythos 5.1
  2. What's new in Claude Fable 5.1
  3. Claude Fable 5.1 & Claude Mythos 5.1 System Card
  4. Benchmarking GPT-6 Astra - Artificial Analysis
  5. OpenAI launches GPT-6 Astra and says welcome to the "AGI era"