Humanity's Last Exam

Every result

• Claude Opus 5.567.7%Reported by the model's maker (Anthropic), Sep 22, 2026

  Anthropic reports 67.7% for claude-opus-5-5 on Humanity's Last Exam with tools at max effort. The result is provider-reported; missing evaluation details limit comparison.

  Setting: Adaptive thinking at max effort, under the announcement’s rule for Opus 5.5 results unless stated otherwise. Numeric reasoning and token budgets are unspecified.

  Limits: Anthropic reports this result; it has not been independently reproduced here. The evaluated checkpoint and serving configuration are not identified. The model name matches the published API identifier, but that does not establish which weights or serving configuration produced the score. The HLE revision, task subset, sample size, harness, prompts, run date, aggregation and confidence interval are not reported in the inspected narrative. Tools were enabled. The tool identities, call limits, runtime and token budgets are not specified; this should not be treated as equivalent to a no-tools evaluation. Max effort is reported, but the numeric reasoning budget is unspecified. Production safeguards could route some tasks to Opus 4.8 or Opus 5. Whether that happened during HLE is unknown, so this is not established as a score for isolated Opus 5.5 weights. Results with the same benchmark name can be viewed together as reported scores. Unmatched settings and missing methods prevent a controlled quality ranking or a conclusion that small differences are meaningful.

  Anthropic

• Claude Sonnet 5.564.5%Reported by the model's maker (Anthropic), Sep 28, 2026

  Anthropic reports 64.5% for claude-sonnet-5-5 on Humanity's Last Exam with tools. The result is provider-reported; missing evaluation details limit comparison.

  Limits: Anthropic reports this result; it has not been independently reproduced here. The evaluated checkpoint and serving configuration are not identified. The model name matches the published API identifier, but that does not establish which weights or serving configuration produced the score. The HLE revision, task subset, sample size, harness, prompts, run date, aggregation and confidence interval are not reported in the inspected narrative. Tools were enabled. The tool identities, call limits, runtime and token budgets are not specified; this should not be treated as equivalent to a no-tools evaluation. Reasoning effort is unspecified; settings from other releases or evaluations cannot fill that gap. The announcement describes deployment safeguards, but whether they affected this evaluation is unknown. Results with the same benchmark name can be viewed together as reported scores. Unmatched settings and missing methods prevent a controlled quality ranking or a conclusion that small differences are meaningful.

  Anthropic

• Claude Opus 563.6%Reported by the model's maker (Anthropic), Sep 22, 2026

  Anthropic reports 63.6% for Claude Opus 5 on Humanity's Last Exam with tools. The result is provider-reported; missing evaluation details limit comparison.

  Limits: Anthropic reports this result; it has not been independently reproduced here. The evaluated checkpoint and serving configuration are not identified. This result refers to the named older release, whose API identifier and current availability are unknown. The HLE revision, task subset, sample size, harness, prompts, run date, aggregation and confidence interval are not reported in the inspected narrative. Tools were enabled. The tool identities, call limits, runtime and token budgets are not specified; this should not be treated as equivalent to a no-tools evaluation. Reasoning effort is unspecified; settings from other releases or evaluations cannot fill that gap. Results with the same benchmark name can be viewed together as reported scores. Unmatched settings and missing methods prevent a controlled quality ranking or a conclusion that small differences are meaningful.

  Anthropic

• Claude Sonnet 554.9%Reported by the model's maker (Anthropic), Sep 28, 2026

  Anthropic reports 54.9% for Claude Sonnet 5 on Humanity's Last Exam with tools. The result is provider-reported; missing evaluation details limit comparison.

  Limits: Anthropic reports this result; it has not been independently reproduced here. The evaluated checkpoint and serving configuration are not identified. This result refers to the named older release, whose API identifier and current availability are unknown. The HLE revision, task subset, sample size, harness, prompts, run date, aggregation and confidence interval are not reported in the inspected narrative. Tools were enabled. The tool identities, call limits, runtime and token budgets are not specified; this should not be treated as equivalent to a no-tools evaluation. Reasoning effort is unspecified; settings from other releases or evaluations cannot fill that gap. Results with the same benchmark name can be viewed together as reported scores. Unmatched settings and missing methods prevent a controlled quality ranking or a conclusion that small differences are meaningful.

  Anthropic