Radar task score: four tests, equal weight. An experimental summary of these tasks, not a universal intelligence rating.
Claude Opus 4.5 (latest)
Radar task score
/ 100
3/4 core tests · not enough evidence
5 benchmark results for this configuration
Scientific reasoning
GPQA Diamond · 25%
Expert-written science questions; strict multiple-choice scoring. Chance accuracy is 25%.
050100%
Claude Opus 4.5 (latest)86.0*%Approx. 95% interval: 82.0–90.1% · Older result
Novel problem solving
Mystery Game Puzzles · 25%
Best-next-move puzzles in an undisclosed game variant. A narrow reasoning test, not general intelligence.
050100%
Claude Opus 4.5 (latest)Not available
Advanced mathematics
FrontierMath Tiers 1-3 · 25%
Private v2 mathematics problems with Python tools. Earlier versions and Tier 4 are excluded.
050100%
Claude Opus 4.5 (latest)34.4*%Approx. 95% interval: 28.9–39.9%
Factual answers
SimpleQA Verified · 25%
Correct factual answers under a guessing prompt, graded by a model. Not a hallucination or abstention score.
050100%
Claude Opus 4.5 (latest)45.7*%Approx. 95% interval: 42.6–48.8%
Competition mathematics
OTIS Mock AIME 2024-2025 · Separate result
Competition problems with model-extracted answers checked against an exact key. Earlier grading procedures are excluded.
050100%
Claude Opus 4.5 (latest)86.1*%Approx. 95% interval: 77.8–94.5% · Older result
Research mathematics
FrontierMath Tier 4 v2 · Separate result
The hardest private FrontierMath problems, corrected v2 set, with Python tools. Separate from Tiers 1-3 in the core score.
050100%
Claude Opus 4.5 (latest)4.9*%Approx. 95% interval: 0.0–11.6%
Visual spatial reasoning
Furniture Assembly · Separate result
Assembly-error detection from photos and manuals, with zoom and Python tools. Uses model-assisted grading; not a general vision score.
050100%
Claude Opus 4.5 (latest)Not available
Bars show reported task scores. Whiskers are approximate 95% intervals where standard errors are available. Different reasoning budgets are not equal-cost tests. Missing results are never zero.
Run it locally
Inference · memory estimate
Claude Opus 4.5 (latest)
Closed weights. No supported local deployment or GPU requirement is published.
API cost
USD / month
Standard text-token estimate. Excludes taxes, caching, batch discounts, reasoning and multimodal surcharges. Provider limits and regional availability may differ.