Claude Fable 5.2 vs GPT-6 Astra
Wins 4 of 6 areas
Coding · Facts · Math · Long documents
Wins 0 of 6 areas
—
Claude Fable 5.2 is the stronger all-rounder.
Both rank among the ten best models we track in math.
Scores updated · 28 tests both models report · How we compare
Where each one wins
Tests won in each of the six areas where both have results. Each piece is one test, so a longer bar means more evidence; grey means the two scored within a point of each other.
- MathCompetition and research-level math20Claude Fable 5.22 of 2 tests
- CodingWriting and fixing software10Claude Fable 5.21 of 1 test
- FactsGetting facts right instead of making them up10Claude Fable 5.21 of 1 test
- Long documentsFinding answers in very long texts10Claude Fable 5.21 of 1 test
- AgentsCarrying out multi-step tasks on its own00Even0 each · 1 tie
- ReasoningHard problems that need careful thinking22Even2 each
Coding, agents, facts and long documents rest on a single test each.
What it costs
Prices per million tokens, roughly 750,000 words. The bars show the cost of a million tokens read plus a million written.
Claude Fable 5.2 costs 20% less for the same work.
The biggest differences
The tests each model wins by the widest margin, up to three each. Scores are out of 100.
Where Claude Fable 5.2 pulls ahead
- Very hard expert questions across many subjectsHumanity's Last Exam+19.6points ahead
- Short factual questions, answered correctlySimpleQA Verified+18.4points ahead
- Code for real scientific research problemsSciCode+11.2points ahead
Every test, side by side
All 28 tests both models report. The winning score is in its model's colour; marks a score checked independently.
ReasoningEven
- ARC-AGI-3GPT-6 Astra by 20.679.399.9+20.6
- Humanity's Last ExamClaude Fable 5.2 by 19.674.354.7+19.6
- CritPtClaude Fable 5.2 by 3.735.431.7+3.7
- ARC-AGI-2GPT-6 Astra by 2.592.595+2.5
MathClaude Fable 5.2
- FrontierMath Tier 4Claude Fable 5.2 by 2.410097.6+2.4
- FrontierMath Tiers 1-3 (v2)Claude Fable 5.2 by 1.495.193.7+1.4
Other results18 tests, not counted
Tests outside the eight areas. They are not counted above: several are summary scores built from other tests, or the same test under another name.
- Program BenchClaude Fable 5.2 by 35.685.650+35.6
- GDPval-AA 2.1Claude Fable 5.2 by 289 rating points18311542+289 rating
- GDPval-AA v2.1 (max effort)Claude Fable 5.2 by 247 rating points18311584+247 rating
- HLE (with tools)Claude Fable 5.2 by 1774.257.2+17
- HLE-DiamondClaude Fable 5.2 by 16.480.263.8+16.4
- Terminal-Bench-Science 0.1Claude Fable 5.2 by 1073.363.3+10
- Terminal-Bench 4.0Claude Fable 5.2 by 9.967.657.7+9.9
- GDP (Surge AI)GPT-6 Astra by 6.824.231+6.8
- PhysicianBench (high effort)Claude Fable 5.2 by 6.861.855+6.8
- CritPt v2 HardClaude Fable 5.2 by 676.570.5+6
- ARC-AGI-3 (Provider Adapter harness, max effort)GPT-6 Astra by 3.794.998.6+3.7
- ARC-AGI-3 (Standard harness, max effort)Claude Fable 5.2 by 365.762.7+3
- HealthBench Professional length-adjustedClaude Fable 5.2 by 367.764.7+3
- HealthBench ProfessionalClaude Fable 5.2 by 1.767.766+1.7
- HealthBenchClaude Fable 5.2 by 1.459.658.2+1.4
- HealthBench length-adjustedClaude Fable 5.2 by 1.359.658.3+1.3
- DeepSWE 1.1tie74.374.1tie
- ARC-AGI-1tie97.597.5tie
Questions people ask
Which is better, Claude Fable 5.2 or GPT-6 Astra?
Claude Fable 5.2 wins four of the six areas where both have results: coding, facts, math and long documents. GPT-6 Astra wins none. They are level on agents and reasoning.
Which is better for coding?
Claude Fable 5.2. It wins the one coding test both models report.
Which is cheaper?
Claude Fable 5.2 costs $8.00 per million input tokens and $40.00 per million output tokens; GPT-6 Astra costs $10.00 and $50.00. That makes Claude Fable 5.2 about 20% cheaper for the same work.
How do you compare the two?
We use the 28 benchmark tests both models have published scores on. The verdict counts the 10 tests in the eight capability areas, and a gap under one point (ten on rating-style scales) counts as a tie. The other 18 are listed but not counted, because several are summary scores or repeat a test. Each score is the one shown on the model's own page.