#3 of 11
0.811HELM Capabilities
A transparent, reproducible evaluation of general language-model capabilities across a curated collection of difficult scenarios.
View full leaderboard →OpenAI model
o3 (2025-04-16) benchmarks
Independent scores collected from compatible, versioned benchmark leaderboards. Each result links back to the evaluation and methodology that produced it.
Measured results
Where this model was evaluated
Scores from different benchmarks measure different capabilities and are intentionally not averaged into one universal rating.
#1 of 10
0.982HELM Safety
A standardized collection of safety evaluations spanning violence, fraud, discrimination, harassment, sexual content and deception.
View full leaderboard →#1 of 4
85.9%Arena-Hard v2.0 Preview
A difficult set of open-ended real-world prompts evaluated with automated judges and style controls.
View full leaderboard →Direct comparisons
Compare common alternatives
Model comparisono3 (2025-04-16)
o3 (2025-04-16)
o3 (2025-04-16)
vs o4-mini (2025-04-16)
Compare only the benchmarks where both configurations have published results.
See comparison →Model comparisono3 (2025-04-16)
vs Claude 4 Sonnet (20250514, extended thinking)
Compare only the benchmarks where both configurations have published results.
See comparison →What the scores mean
A benchmark score describes performance under one specific evaluation setup. It does not guarantee the same ordering for a different task, prompt, model version or deployment configuration.
Source policy
Shareof.ai preserves the published model name, benchmark version and source link. Provider-reported claims are not presented as independent results.
Read our methodology →Models are measured. Brands should be too.
See if AI recommends your company.
Run a live sample and inspect the competitors and sources appearing in the answers.
Check my AI visibility →