CategoryWeight: 1.0x

Long Reasoning

Multi-step logic puzzles, extended chain-of-thought, and complex analytical reasoning tasks requiring sustained coherence over many steps.

Best Score

0.0

Avg Score

0.0

Tests

Performance Over Time — All Models

Model Rankings

Claude Sonnet 4.6

Category score

View

70.7BEST

Tokens56.0k

Total56.0k

Claude Opus 4.8

Category score

View

70.3-0.4 pts

Tokens19.1k

Total19.1k

Grok

Category score

View

68.4-2.3 pts

Tokens67.9k

Total67.9k

GPT-5.5

Category score

View

63.4-7.3 pts

Tokens38.3k

Total38.3k

Rank	Model	Score	Tokens	vs. Best	Details
1	Claude Sonnet 4.6	70.7	56.0k	BEST	View
2	Claude Opus 4.8	70.3	19.1k	-0.4 pts	View
3	Grok	68.4	67.9k	-2.3 pts	View
4	GPT-5.5	63.4	38.3k	-7.3 pts	View

Test Breakdown

Multi-step Logic Puzzle

Complex optimization with 8+ constraints across multiple variables

Claude Sonnet 4.6

70.7

Claude Opus 4.8

70.3

Grok

68.4

GPT-5.5

63.4

Legal Reasoning Chain

Contract dispute analysis requiring multi-party obligation tracking

Claude Sonnet 4.6

70.7

Claude Opus 4.8

70.3

Grok

68.4

GPT-5.5

63.4

Mathematical Proof

Prove divisibility properties using induction and modular arithmetic

Claude Sonnet 4.6

70.7

Claude Opus 4.8

70.3

Grok

68.4

GPT-5.5

63.4