RLTasksby ulam.ai
Models / Anthropic

Anthropic / MODEL EVIDENCE

Claude Opus 5.0

1 benchmarks · 8 observations · 1 sources

Scores at a glance

1 benchmarks · large dot: Claude Opus 5.0; small dots: closest models where reported

TasteVal ↗0.58 ×
0.012.3

Opus 5.0 (low) · TasteVal · Experimental compute multiplier · October 2026 paper comparison ↗

7 more reported results

0.92 × · Opus 5.0 (low) · TasteVal · Normalized final-performance multiplier · October 2026 paper comparison · TasteVal: Measuring the Experimental Research Taste of AI Systems Against Human Experts ↗

2.11 × · Opus 5.0 (max, Super-Think) · TasteVal · Experimental compute multiplier · October 2026 paper comparison · TasteVal: Measuring the Experimental Research Taste of AI Systems Against Human Experts ↗

1.29 × · Opus 5.0 (max) · TasteVal · Experimental compute multiplier · October 2026 paper comparison · TasteVal: Measuring the Experimental Research Taste of AI Systems Against Human Experts ↗

0.97 × · Opus 5.0 (medium) · TasteVal · Experimental compute multiplier · October 2026 paper comparison · TasteVal: Measuring the Experimental Research Taste of AI Systems Against Human Experts ↗

1.29 × · Opus 5.0 (max, Super-Think) · TasteVal · Normalized final-performance multiplier · October 2026 paper comparison · TasteVal: Measuring the Experimental Research Taste of AI Systems Against Human Experts ↗

1.11 × · Opus 5.0 (max) · TasteVal · Normalized final-performance multiplier · October 2026 paper comparison · TasteVal: Measuring the Experimental Research Taste of AI Systems Against Human Experts ↗

1.07 × · Opus 5.0 (medium) · TasteVal · Normalized final-performance multiplier · October 2026 paper comparison · TasteVal: Measuring the Experimental Research Taste of AI Systems Against Human Experts ↗

Closest models by score

Similar positions on shared benchmark leaderboards. Smaller gaps are closer.

4.8 percentile-point gap

1 of 1 shared benchmark comparisons

How similarity is measured

Mean absolute gap in percentile positions on shared benchmark settings; lower is closer. At least 60% overlap is required, with a minimum of three when available. Independent evaluations are used when there are at least three; otherwise published report comparisons are used. Repeated model observations within one setting are averaged before ranking. Writing-style traits and composite indices are excluded. This describes similarity, not an overall quality ranking.

Filter comparisons to this model →

Sources

Every plotted observation links to its evaluation. Reported configurations, benchmark versions and task subsets can differ; similarity describes shared scores, not training lift.

  1. TasteVal: Measuring the Experimental Research Taste of AI Systems Against Human Experts — published result table ↗

    Jaffe, Oliver, Sherburn, Dane · 2026

Names used in source tables

Opus 5.0 (low) · Opus 5.0 (max) · Opus 5.0 (max, Super-Think) · Opus 5.0 (medium)