Claude Opus 5.5

Leaderboards

Where Claude Opus 5.5 sits among the top 20 in each category.

By task

Scored from votes on each kind of prompt alone, on the same 1,000-point scale.

Test results

Every test Claude Opus 5.5 has a published result on, out of 100%, with its rank among the models that took it.

• SWE-Bench Pro

  #1 / 61

  Harder, longer software engineering tasks from real codebases.

  89.9%

  Self-reported · In the coding score

• DeepSWE 1.1

  #10 / 50

  DeepSWE 1.1 evaluates software engineering agents using the mini-swe-agent harness.

  74.2%

  Self-reported · In the coding score

• SWE-bench Multilingual

  #1 / 49

  A multilingual benchmark for issue resolving in software engineering that covers Java, TypeScript, JavaScript, Go, Rust, C, and C++. Contains 1,632 high-quality instances carefully annotated from 2,456 candidates by 68 expert annotators, designed to evaluate Large Language Models across diverse software ecosystems beyond Python.

  93.9%

  Self-reported

• Terminal-Bench 4.0

  #2 / 28

  Terminal-Bench 4.0 evaluates AI agents on real-world work in terminal and command-line containerized environments across 66 tasks, with emphasis on science-adjacent and frontier engineering problems. Relative to earlier Terminal-Bench releases, 4.0 increases timeouts and adaptively raises RAM/CPU on selected tasks to reduce harness and resource confounds.

  66.4%

  Self-reported · In the coding score· In the agents score

• AutomationBench

  #10 / 24

  AutomationBench is a tool-use benchmark that evaluates AI agents on automating real-world workflows, testing their ability to orchestrate tools and complete multi-step automation tasks.

  40.0%

  Self-reported · In the agents score

• FrontierCode 1.1

  #1 / 23

  FrontierCode 1.1 evaluates whether coding-agent changes are mergeable, using unit tests, maintainer-defined rubrics, and verifiers. Runs flagged for unfair internet use receive a zero score.

  54.4%

  Self-reported · In the coding score

• HealthBench Professional

  #3 / 19

  HealthBench Professional evaluates model capability and safety for clinician use cases using real clinician-style chats and physician-authored grading rubrics.

  65.6%

  Self-reported

• Program Bench

  #1 / 18

  Program Bench evaluates code-generation agents by asking them to recreate a program's behavior from only a compiled binary and documentation. It spans 200 tasks from small CLI tools to large systems such as FFmpeg and SQLite, with submissions judged against more than 248,000 fuzz-generated behavioral tests.

  91.2%

  Self-reported · In the coding score

• OSWorld 2.0

  #1 / 17

  OSWorld 2.0 is a benchmark of 108 long-horizon, real-world computer-use workflows spanning everyday and professional tasks. Each task is an end-to-end workflow that takes human users a median of about 1.6 hours, scored with a binary-completion metric, and targets challenges such as dynamic environments, cross-source reasoning, and implicit-state inference.

  81.8%

  Self-reported

• Legal Agent Benchmark

  #4 / 15

  The Legal Agent Benchmark (LAB) is Harvey's open-source benchmark for evaluating AI agents on complex, long-horizon legal work. Tasks are scored under an all-pass standard against expert-curated rubrics, where a task passes only if every required rubric criterion (facts, conclusions, citations, structure, and analytical moves) passes.

  8.3%

  Self-reported

• HealthBench

  #3 / 14

  An open-source benchmark for measuring performance and safety of large language models in healthcare, consisting of 5,000 multi-turn conversations evaluated by 262 physicians using 48,562 unique rubric criteria across health contexts and behavioral dimensions

  60.6%

  Self-reported

• Humanity's Last Exam (no tools, text-only)

  #1 / 14

  Text-only Humanity's Last Exam variant evaluated without tool use.

  64.4%

  Self-reported

• OfficeQA Pro

  #3 / 12

  OfficeQA Pro evaluates AI models on professional knowledge-work questions and tasks drawn from real office workflows, including document analysis, spreadsheet reasoning, and information synthesis across business domains.

  67.7%

  Self-reported

• Humanity's Last Exam (with tools, text-only)

  #1 / 10

  Text-only Humanity's Last Exam variant evaluated with tool use enabled.

  67.7%

  Self-reported

• Global-MMLU

  #1 / 8

  A comprehensive multilingual benchmark covering 42 languages that addresses cultural and linguistic biases in evaluation, with improved translation quality and culturally sensitive question subsets.

  94.3%

  Self-reported

• BenchCAD

  #2 / 7

  BenchCAD is a benchmark for programmatic CAD reasoning built from 17,900 execution-verified CadQuery programs spanning 106 industrial part families, roughly half anchored to real ISO, DIN, EN, ASME, and IEC specification tables. It decomposes CAD capability into matched tasks; the Vision2Code task requires models to generate CadQuery code from multi-view renders, scored by voxel IoU against the reference geometry.

  73.0%

  Self-reported

• BenchCAD (with Python tool)

  #2 / 7

  BenchCAD variant evaluated with access to a Python tool for programmatic CAD reasoning.

  96.2%

  Self-reported

• BioMysteryBench

  #2 / 7

  BioMysteryBench evaluates a model's ability to reason through challenging molecular biology problems, reporting performance on a hard subset and on the subset of problems solved by human experts.

  89.3%

  Self-reported

• SWE-Bench Multimodal

  #1 / 7

  SWE-Bench Multimodal extends SWE-Bench to evaluate language models on software engineering tasks that involve visual inputs such as screenshots, UI mockups, and diagrams alongside code understanding.

  61.4%

  Self-reported

• Terminal-Bench-Science 0.1

  #3 / 7

  Terminal-Bench-Science 0.1 is a Stanford-led community benchmark of 70 tasks drawn from scientific research workflows across the life, physical, earth, mathematical, and engineering sciences. Tasks are authored and reviewed by scientists and researchers; scores are typically reported as accuracy with relatively large standard error due to strongly bimodal per-task outcomes.

  58.7%

  Self-reported

• Chartography

  #2 / 6

  Chartography evaluates professional chart understanding with 100 expert-written tasks drawn from real charts across science, engineering, finance, manufacturing, supply chain, and healthcare.

  89.0%

  Self-reported

• AA-Briefcase v1.1

  #1 / 5

  AA-Briefcase v1.1 evaluates multi-hour professional office work and reports Elo scores. Kept separate from unversioned AA-Briefcase results.

  60.7%

  Self-reported

• ArXivMath

  #1 / 5

  ArXivMath is a final-answer benchmark of research-level mathematics maintained by MathArena. Problems are extracted monthly from recent arXiv paper abstracts, then filtered through automated and manual checks to ensure they are self-contained, non-trivial, and verifiable. Because problems are drawn from active research, the benchmark is more realistic and more closely connected to mathematical research than contest or olympiad benchmarks.

  91.2%

  Self-reported

• FrontierSWE V2

  #1 / 5

  FrontierSWE V2 evaluates agents on open-ended technical projects and reports mean@5. Kept separate from FrontierSWE V1 dominance-score results, which are not comparable.

  62.3%

  Self-reported

• GDPval-AA 2.1

  #1 / 5

  Version 2.1 of the Artificial Analysis GDPval evaluation of professional knowledge work, reported as an Elo rating.

  61.5%

  Self-reported

• Toolathlon-Verified

  #1 / 5

  Verified Toolathlon evaluation of agent tool-use capability, as reported in the MiMo-V2.6 release.

  77.8%

  Self-reported

• CursorBench 4.0

  #1 / 3

  CursorBench 4.0 evaluates coding agents on long-horizon coding tasks from real Cursor sessions. Version 4.0 adds longer-horizon tasks; scores are not comparable with CursorBench 3.2.

  57.8%

  Self-reported

• LatchBio SingleCellBench

  #1 / 3

  LatchBio SingleCellBench tests analysis of single-cell RNA sequencing data across problems spanning standard workflows such as labeling cell types, finding differentially expressed genes, and correcting batch effects.

  61.2%

  Self-reported

• LatchBio SpatialBench Verified

  #2 / 3

  LatchBio SpatialBench Verified tests analysis of spatial transcriptomics data across externally validated problems, requiring the model to answer biological questions about the sample from analysis results.

  72.0%

  Self-reported

• MILU

  #1 / 3

  Multi-task Indic Language Understanding (MILU) covers 11 languages—10 Indic languages (Bengali, Gujarati, Hindi, Kannada, Malayalam, Marathi, Odia, Punjabi, Tamil, and Telugu) and English—and tests culturally grounded knowledge comprehension.

  93.1%

  Self-reported

• OfficeQA

  #1 / 3

  OfficeQA is a Databricks public benchmark that evaluates end-to-end grounded reasoning over a large corpus of historical U.S. Treasury Bulletin documents. Models must locate relevant tables across the corpus and perform precise numerical reasoning over them.

  78.9%

  Self-reported