New: Claude Fable 5.2 eval report is here.Read the report ↗

← Leaderboard

Compare models

Find the right model, side by side.

Updated: Nov 9, 2026Data as of: Nov 9, 2026

First model

Second model

Model specifications

Specifications for Claude Fable 5.2 and GPT-6 Astra. Prices and performance use the selected provider offering.
Specifications	Claude Fable 5.2Anthropic	GPT-6 AstraOpenAI
Provider
Context sizetokens	1,000,000	1,000,000
Cutoff date	2026-06	2026-04-30
Input costUSD / 1M tokens	$16.00	$10.00
Output costUSD / 1M tokens	$80.00	$50.00
Max outputtokens	128,000	128,000
LatencyTTFT · seconds	Not reported	365.127
Speedtokens / sec	Not reported	70.564

Benchmark performance

18 shared benchmark comparisons. Versions and setups matched.

Claude Fable 5.2GPT-6 Astra

AA-Briefcase

1.1 / rubric pass rate ↗Tools enabled

Claude Fable 5.2: 66.1%
GPT-6 Astra: 52.0%

AutomationBench-AA

1.0.6 / AA private 657 ↗Tools enabled

Claude Fable 5.2: 76.4%
GPT-6 Astra: 68.5%

Chartography

Anthropic implementation / Gemini 3.5 Flash grader ↗No tools

Claude Fable 5.2: 73.4%
GPT-6 Astra: 71.0%

FrontierCode

2.0 / Cognition native agents ↗Tools enabled

Claude Fable 5.2: 49.1%
GPT-6 Astra: 42.8%

FrontierSWE

2 / Proximal harness ↗Tools enabled

Claude Fable 5.2: 69.9%
GPT-6 Astra: 65.5%

Harvey LAB-AA

1.1 / private 120 / hallucination-gated ↗Tools enabled · Stirrup

Claude Fable 5.2: 7.2%
GPT-6 Astra: 8.6%

HLE-Diamond

1,000 questions / no tools ↗No tools

Claude Fable 5.2: 80.2%
GPT-6 Astra: 63.8%

LiveBench: Agentic

2.0 / 2026-11-06 ↗

Claude Fable 5.2: 82.9%
GPT-6 Astra: 75.6%

LiveBench: Browser Use

2.0 / 2026-11-06 ↗

Claude Fable 5.2: 86.9%
GPT-6 Astra: 91.3%

LiveBench: Coding

2.0 / 2026-11-06 ↗

Claude Fable 5.2: 79.2%
GPT-6 Astra: 87.5%

LiveBench: Data Analysis

2.0 / 2026-11-06 ↗

Claude Fable 5.2: 81.5%
GPT-6 Astra: 83.0%

LiveBench: Design

2.0 / 2026-11-06 ↗

Claude Fable 5.2: 85.1%
GPT-6 Astra: 80.0%

LiveBench: Reasoning

2.0 / 2026-11-06 ↗

Claude Fable 5.2: 88.8%
GPT-6 Astra: 92.7%

LiveBench: Research

2.0 / 2026-11-06 ↗

Claude Fable 5.2: 94.5%
GPT-6 Astra: 90.6%

OSWorld

2.1 / offline 82 / partial credit ↗Tools enabled

Claude Fable 5.2: 89.3%
GPT-6 Astra: 79.5%

Remote Labor Index

240 freelance projects ↗Tools enabled

Claude Fable 5.2: 41.3%
GPT-6 Astra: 20.8%

RoboDojo Sim

Anthropic internal simulation ↗

Claude Fable 5.2: 64.9%
GPT-6 Astra: 29.8%

Terminal-Bench-Science

0.1 / Claude Code / offline ↗Tools enabled

Claude Fable 5.2: 74.6%
GPT-6 Astra: 68.1%

Model comparisons