Shared benchmarks

41
Shared benchmarks for GPT-5.4 and GPT-5.5
BenchmarkGPT-5.4GPT-5.5
  • 52.4%
  • 55.1%
Evaluation details for APEX-Agents Original

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on APEX-Agents Original

GPT-5.5

Record for GPT-5.5 on APEX-Agents Original
  • 1.5%low
  • 5.3%medium
  • 10.1%high
  • 11.6%xhigh
  • 0.2%
  • 2.6%low
  • 9.0%medium
  • 12.8%high
  • 16.6%xhigh
  • 1.1%
Evaluation details for AutomationBench 1.0.6

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on AutomationBench 1.0.6
Record for GPT-5.4 on AutomationBench 1.0.6
Record for GPT-5.4 on AutomationBench 1.0.6
Record for GPT-5.4 on AutomationBench 1.0.6
Record for GPT-5.4 on AutomationBench 1.0.6

GPT-5.5

Record for GPT-5.5 on AutomationBench 1.0.6
Record for GPT-5.5 on AutomationBench 1.0.6
Record for GPT-5.5 on AutomationBench 1.0.6
Record for GPT-5.5 on AutomationBench 1.0.6
Record for GPT-5.5 on AutomationBench 1.0.6
  • 73.3%
  • 77.7%
Evaluation details for BrowseComp 130-question subset — Mercor web research agent

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

GPT-5.5

  • 82.7%
  • 84.4%
Evaluation details for BrowseComp

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on BrowseComp BrowseComp

GPT-5.5

Record for GPT-5.5 on BrowseComp BrowseComp
  • 29.6%
  • 31.0%
Evaluation details for Chartography October 2026 leaderboard

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on Chartography October 2026 leaderboard

GPT-5.5

Record for GPT-5.5 on Chartography October 2026 leaderboard
  • 93.4%
  • 94.6%
Evaluation details for CharXiv Descriptive and reasoning — Mercor single-shot

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

  • 49.0%
  • 66.1%
Evaluation details for DeepSWE 1.1 — Mercor mini-swe-agent, 500 steps

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on DeepSWE 1.1 — Mercor mini-swe-agent, 500 steps

GPT-5.5

Record for GPT-5.5 on DeepSWE 1.1 — Mercor mini-swe-agent, 500 steps
  • 83.0%
  • 84.9%
Evaluation details for GDPval

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on GDPval GDPval

GPT-5.5

Record for GPT-5.5 on GDPval GDPval
  • 92.8%
  • 93.6%
Evaluation details for GPQA Diamond

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on Graduate-Level Google-Proof Q&A Diamond

GPT-5.5

Record for GPT-5.5 on Graduate-Level Google-Proof Q&A Diamond
  • 74.7%none
  • 84.8%low
  • 88.9%medium
  • 89.9%high
  • 93.3%xhigh
  • 77.3%none
  • 90.7%low
Evaluation details for GPQA Diamond — Epoch AI harness

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on Graduate-Level Google-Proof Q&A Diamond — Epoch AI harness
Record for GPT-5.4 on Graduate-Level Google-Proof Q&A Diamond — Epoch AI harness
Record for GPT-5.4 on Graduate-Level Google-Proof Q&A Diamond — Epoch AI harness
Record for GPT-5.4 on Graduate-Level Google-Proof Q&A Diamond — Epoch AI harness
Record for GPT-5.4 on Graduate-Level Google-Proof Q&A Diamond — Epoch AI harness

GPT-5.5

Record for GPT-5.5 on Graduate-Level Google-Proof Q&A Diamond — Epoch AI harness
Record for GPT-5.5 on Graduate-Level Google-Proof Q&A Diamond — Epoch AI harness
  • 93.2%
  • 94.6%
Evaluation details for GPQA Diamond — Mercor single-shot

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on Graduate-Level Google-Proof Q&A Diamond — Mercor single-shot

GPT-5.5

Record for GPT-5.5 on Graduate-Level Google-Proof Q&A Diamond — Mercor single-shot
  • 91.7%
  • 93.2%
Evaluation details for GPQA Diamond — Vals AI harness

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on Graduate-Level Google-Proof Q&A Diamond — Vals AI harness

GPT-5.5

Record for GPT-5.5 on Graduate-Level Google-Proof Q&A Diamond — Vals AI harness
  • 0.0%
  • 3.8%
Evaluation details for HLAB October 2026 evaluation

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on Harvey's Legal Agent Benchmark October 2026 evaluation

GPT-5.5

Record for GPT-5.5 on Harvey's Legal Agent Benchmark October 2026 evaluation
  • 39.8%
  • 41.4%
Evaluation details for Humanity's Last Exam Full set — No tools

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on Humanity's Last Exam Full set — No tools

GPT-5.5

Record for GPT-5.5 on Humanity's Last Exam Full set — No tools
  • 52.1%
  • 52.2%
Evaluation details for Humanity's Last Exam Full set — With tools

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on Humanity's Last Exam Full set — With tools

GPT-5.5

Record for GPT-5.5 on Humanity's Last Exam Full set — With tools
  • 44.1%
  • 46.5%
Evaluation details for Humanity's Last Exam Text only — Mercor single-shot

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on Humanity's Last Exam Text only — Mercor single-shot

GPT-5.5

Record for GPT-5.5 on Humanity's Last Exam Text only — Mercor single-shot
  • 70.6%
  • 75.3%
Evaluation details for MCP Atlas Public, April 2026 update

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on MCP Atlas Public, April 2026 update

GPT-5.5

Record for GPT-5.5 on MCP Atlas Public, April 2026 update
  • 74.2%
  • 76.7%
Evaluation details for MedXpertQA MM — Mercor single-shot

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on MedXpertQA MM — Mercor single-shot

GPT-5.5

Record for GPT-5.5 on MedXpertQA MM — Mercor single-shot
  • 88.2%
  • 88.9%
Evaluation details for MMLU-Pro Original — Mercor single-shot

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on MMLU-Pro Original — Mercor single-shot

GPT-5.5

Record for GPT-5.5 on MMLU-Pro Original — Mercor single-shot
  • 87.5%
  • 88.1%
Evaluation details for MMLU-Pro Original — Vals AI harness

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on MMLU-Pro Original — Vals AI harness

GPT-5.5

Record for GPT-5.5 on MMLU-Pro Original — Vals AI harness
  • 80.8%
  • 81.8%
Evaluation details for MMMU-Pro 1,730 questions — Mercor single-shot

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on MMMU-Pro 1,730 questions — Mercor single-shot

GPT-5.5

Record for GPT-5.5 on MMMU-Pro 1,730 questions — Mercor single-shot
  • 81.2%
  • 81.2%
Evaluation details for MMMU-Pro No tools

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on MMMU-Pro No tools

GPT-5.5

Record for GPT-5.5 on MMMU-Pro No tools
  • 87.5%
  • 88.3%
Evaluation details for MMMU-Pro Vals AI harness

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on MMMU-Pro Vals AI harness

GPT-5.5

Record for GPT-5.5 on MMMU-Pro Vals AI harness
  • 82.1%
  • 83.2%
Evaluation details for MMMU-Pro With tools

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on MMMU-Pro With tools

GPT-5.5

Record for GPT-5.5 on MMMU-Pro With tools
  • 75.0%
  • 78.7%
Evaluation details for OSWorld Verified

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on OSWorld Verified

GPT-5.5

Record for GPT-5.5 on OSWorld Verified
  • 19.0%
  • 27.2%
Evaluation details for PostTrainBench 1.1 — Codex CLI

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on PostTrainBench 1.1 — Codex CLI

GPT-5.5

Record for GPT-5.5 on PostTrainBench 1.1 — Codex CLI
  • 17.8%
  • 23.9%
Evaluation details for PostTrainBench 1.2 — Codex CLI

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on PostTrainBench 1.2 — Codex CLI

GPT-5.5

Record for GPT-5.5 on PostTrainBench 1.2 — Codex CLI
  • 41.6%
  • 55.2%
Evaluation details for Riemann-bench October 2026 leaderboard

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on Riemann-bench October 2026 leaderboard

GPT-5.5

Record for GPT-5.5 on Riemann-bench October 2026 leaderboard
  • 45.1%
  • 63.0%
Evaluation details for SimpleQA Verified Original — Epoch AI harness

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

  • 37.1%
  • 45.2%
Evaluation details for SWE Atlas Codebase QnA — Mercor mini-swe-agent, 250 steps

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

  • 40.8%
  • 45.4%
Evaluation details for SWE Atlas Codebase QnA — Public; Codex

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

GPT-5.5

  • 44.3%
  • 44.8%
Evaluation details for SWE Atlas Refactoring — Codex

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on SWE Atlas Refactoring — Codex

GPT-5.5

Record for GPT-5.5 on SWE Atlas Refactoring — Codex
  • 77.0%
  • 77.7%
Evaluation details for SWE-bench Multilingual — Mercor mini-swe-agent, 1000 steps

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

  • 78.1%
  • 78.8%
Evaluation details for SWE-bench Verified — Mercor mini-swe-agent, 1000 steps

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

GPT-5.5

  • 78.2%
  • 82.6%
Evaluation details for SWE-bench Verified — Vals AI mini-swe-agent

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on SWE-bench Verified — Vals AI mini-swe-agent

GPT-5.5

Record for GPT-5.5 on SWE-bench Verified — Vals AI mini-swe-agent
  • 75.1%
  • 82.7%
Evaluation details for Terminal-Bench 2.0

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on Terminal-Bench 2.0

GPT-5.5

Record for GPT-5.5 on Terminal-Bench 2.0
  • 58.4%
  • 73.2%
Evaluation details for Terminal-Bench 2.0 — Vals AI Terminus 2

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on Terminal-Bench 2.0 — Vals AI Terminus 2

GPT-5.5

Record for GPT-5.5 on Terminal-Bench 2.0 — Vals AI Terminus 2
  • 77.2%
  • 83.1%
Evaluation details for Terminal-Bench 2.1 — Mercor Terminus 2

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

GPT-5.4

Record for GPT-5.4 on Terminal-Bench 2.1 — Mercor Terminus 2

GPT-5.5

Record for GPT-5.5 on Terminal-Bench 2.1 — Mercor Terminus 2
  • 48.5%
  • 58.2%
Evaluation details for Vibe Code Bench 1.1 — Codex, October 2026 evaluation

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

  • 67.4%
  • 69.8%
Evaluation details for Vibe Code Bench 1.1 — OpenHands, October 2026 evaluation

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

  • 39.4%
  • 44.6%
Evaluation details for τ³-bench τ³ — Banking, AllTools retrieval, v1.0.1 grading

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.