GPT-5

Leaderboards

Where GPT-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 GPT-5 has a published result on, out of 100%, with its rank among the models that took it.

• GPQA

  #63 / 256

  Graduate-level science questions written to be hard to look up.

  85.7%

  Self-reported

• AIME 2025

  #22 / 123

  Competition maths problems from the 2025 American Invitational Mathematics Exam.

  94.6%

  Self-reported

• SWE-Bench Verified

  #40 / 120

  Fixing real bugs in real GitHub projects, checked by the projects' own tests.

  74.9%

  Self-reported

• Humanity's Last Exam

  #63 / 108

  Humanity's Last Exam (HLE) is a multi-modal academic benchmark with 2,500 questions across mathematics, humanities, and natural sciences, designed to test LLM capabilities at the frontier of human knowledge with unambiguous, verifiable solutions

  24.8%

  Self-reported · In the text score· In the vision score

• MMLU

  #1 / 103

  Massive Multitask Language Understanding benchmark testing knowledge across 57 diverse subjects including STEM, humanities, social sciences, and professional domains

  92.5%

  Self-reported

• MMMU-Pro

  #20 / 73

  College-level questions that need both an image and text to answer.

  78.4%

  Self-reported · In the vision score

• MATH

  #13 / 71

  MATH dataset contains 12,500 challenging competition mathematics problems from AMC 10, AMC 12, AIME, and other mathematics competitions. Each problem includes full step-by-step solutions and spans multiple difficulty levels (1-5) across seven mathematical subjects including Prealgebra, Algebra, Number Theory, Counting and Probability, Geometry, Intermediate Algebra, and Precalculus.

  84.7%

  Self-reported

• BrowseComp

  #45 / 68

  Finding hard-to-locate facts by browsing the web.

  54.9%

  Self-reported · In the agents score

• HumanEval

  #4 / 66

  A benchmark that measures functional correctness for synthesizing programs from docstrings, consisting of 164 original programming problems assessing language comprehension, algorithms, and simple mathematics

  93.4%

  Self-reported

• MMMU

  #5 / 66

  MMMU (Massive Multi-discipline Multimodal Understanding) is a benchmark designed to evaluate multimodal models on college-level subject knowledge and deliberate reasoning. Contains 11.5K meticulously collected multimodal questions from college exams, quizzes, and textbooks, covering six core disciplines: Art & Design, Business, Science, Health & Medicine, Humanities & Social Science, and Tech & Engineering across 30 subjects and 183 subfields.

  84.2%

  Self-reported

• CharXiv-R

  #25 / 60

  Reading and reasoning about charts from scientific papers.

  81.1%

  Self-reported

• Multi-Challenge

  #4 / 41

  MultiChallenge is a realistic multi-turn conversation evaluation benchmark that challenges frontier LLMs across four key categories: instruction retention (maintaining instructions throughout conversations), inference memory (recalling and connecting details from previous turns), reliable versioned editing (adapting to evolving instructions during collaborative editing), and self-coherence (avoiding contradictions in responses). The benchmark evaluates models on sustained, contextually complex dialogues across diverse topics including travel planning, technical documentation, and professional communication.

  69.6%

  Self-reported

• Tau2 Telecom

  #9 / 36

  τ²-Bench telecom domain evaluates conversational agents in a dual-control environment modeled as a Dec-POMDP, where both agent and user use tools in shared telecommunications troubleshooting scenarios that test coordination and communication capabilities.

  96.7%

  Self-reported

• HMMT 2025

  #12 / 33

  Harvard-MIT Mathematics Tournament 2025 - A prestigious student-organized mathematics competition for high school students featuring two tournaments (November 2025 at MIT and February 2026 at Harvard) with individual tests, team rounds, and guts rounds

  93.3%

  Self-reported

• ERQA

  #2 / 29

  Embodied Reasoning Question Answering benchmark consisting of 400 multiple-choice visual questions across spatial reasoning, trajectory reasoning, action reasoning, state estimation, and multi-view reasoning for evaluating AI capabilities in physical world interactions

  2000.0%

  Verified

• VideoMMMU

  #7 / 28

  Video-MMMU evaluates Large Multimodal Models' ability to acquire knowledge from expert-level professional videos across six disciplines through three cognitive stages: perception, comprehension, and adaptation. Contains 300 videos and 900 human-annotated questions spanning Art, Business, Science, Medicine, Humanities, and Engineering.

  84.6%

  Self-reported

• Tau2 Retail

  #7 / 27

  τ²-bench retail domain evaluates conversational AI agents in customer service scenarios within a dual-control environment where both agent and user can interact with tools. Tests tool-agent-user interaction, rule adherence, and task consistency in retail customer support contexts.

  81.1%

  Self-reported

• Tau2 Airline

  #12 / 25

  TAU2 airline domain benchmark for evaluating conversational agents in dual-control environments where both AI agents and users interact with tools in airline customer service scenarios. Tests agent coordination, communication, and ability to guide user actions in tasks like flight booking, modifications, cancellations, and refunds.

  62.6%

  Self-reported

• HealthBench Hard

  #23 / 24

  A challenging variation of HealthBench that evaluates large language models' performance and safety in healthcare through 5,000 multi-turn conversations with particularly rigorous evaluation criteria validated by 262 physicians from 60 countries

  1.6%

  Self-reported

• Aider-Polyglot

  #1 / 22

  A coding benchmark that evaluates LLMs on 225 challenging Exercism programming exercises across C++, Go, Java, JavaScript, Python, and Rust. Models receive two attempts to solve each problem, with test error feedback provided after the first attempt if it fails. The benchmark measures both initial problem-solving ability and capacity to edit code based on error feedback, providing an end-to-end evaluation of code generation and editing capabilities across multiple programming languages.

  88.0%

  Self-reported

• FrontierMath

  #11 / 17

  A benchmark of hundreds of original, exceptionally challenging mathematics problems crafted and vetted by expert mathematicians, covering most major branches of modern mathematics from number theory and real analysis to algebraic geometry and category theory.

  26.3%

  Self-reported

• Graphwalks BFS <128k

  #4 / 13

  A graph reasoning benchmark that evaluates language models' ability to perform breadth-first search (BFS) operations on graphs with context length under 128k tokens, returning nodes reachable at specified depths.

  78.3%

  Self-reported

• Graphwalks parents <128k

  #4 / 12

  A graph reasoning benchmark that evaluates language models' ability to find parent nodes in graphs with context length under 128k tokens, requiring understanding of graph structure and edge relationships.

  73.3%

  Self-reported

• VideoMME w sub.

  #6 / 12

  The first-ever comprehensive evaluation benchmark of Multi-modal LLMs in Video analysis. Features 900 videos (254 hours) with 2,700 question-answer pairs covering 6 primary visual domains and 30 subfields. Evaluates temporal understanding across short (11 seconds) to long (1 hour) videos with multi-modal inputs including video frames, subtitles, and audio.

  86.7%

  Self-reported

• COLLIE

  #1 / 11

  COLLIE is a grammar-based framework for systematic construction of constrained text generation tasks. It allows specification of rich, compositional constraints across diverse generation levels and modeling challenges including language understanding, logical reasoning, and semantic planning. The COLLIE-v1 dataset contains 2,080 instances across 13 constraint structures.

  99.0%

  Self-reported

• OpenAI-MRCR: 2 needle 128k

  #1 / 9

  Multi-round Co-reference Resolution (MRCR) benchmark for evaluating an LLM's ability to distinguish between multiple needles hidden in long context. Models are given a long, multi-turn synthetic conversation and must retrieve a specific instance of a repeated request, requiring reasoning and disambiguation skills beyond simple retrieval.

  95.2%

  Self-reported

• Internal API instruction following (hard)

  #1 / 7

  Internal API instruction following (hard) benchmark - specific documentation not found in official sources

  64.0%

  Self-reported

• SWE-Lancer (IC-Diamond subset)

  #1 / 6

  SWE-Lancer (IC-Diamond subset) is a benchmark of real-world freelance software engineering tasks from Upwork, ranging from $50 bug fixes to $32,000 feature implementations. It evaluates AI models on independent engineering tasks using end-to-end tests triple-verified by experienced software engineers, and includes managerial tasks where models choose between technical implementation proposals.

  100.0%

  Self-reported

• BrowseComp Long Context 128k

  #2 / 5

  A challenging benchmark for evaluating web browsing agents' ability to persistently navigate the internet and find hard-to-locate, entangled information. Comprises 1,266 questions requiring strategic reasoning, creative search, and interpretation of retrieved content, with short and easily verifiable answers.

  90.0%

  Self-reported

• ERQA (default)

  #3 / 3

  ERQA (Embodied Reasoning Question Answering) is a 400-question multimodal multiple-choice benchmark introduced with Google DeepMind's Gemini Robotics report. Questions span spatial, trajectory, action, state, multi-view, pointing, task, and other reasoning categories aimed at physical-world understanding.

  3375.0%

  Unverified

• FActScore

  #3 / 3

  A fine-grained atomic evaluation metric for factual precision in long-form text generation that breaks generated text into atomic facts and computes the percentage supported by reliable knowledge sources, with automated assessment using retrieval and language models

  1.0%

  Self-reported

• BrowseComp Long Context 256k

  #2 / 2

  BrowseComp is a benchmark for measuring the ability of agents to browse the web, comprising 1,266 questions that require persistently navigating the internet in search of hard-to-find, entangled information. Despite the difficulty of the questions, BrowseComp is simple and easy-to-use, as predicted answers are short and easily verifiable against reference answers. The benchmark focuses on questions where answers are obscure, time-invariant, and well-supported by evidence scattered across the open web.

  88.8%

  Unverified

• LongFact Concepts

  #2 / 2

  LongFact is a benchmark for evaluating long-form factuality in large language models. It comprises 2,280 fact-seeking prompts spanning 38 topics, designed to test a model's ability to generate accurate, long-form responses. The benchmark uses SAFE (Search-Augmented Factuality Evaluator) to evaluate factual accuracy.

  0.7%

  Unverified

• LongFact Objects

  #2 / 2

  LongFact is a benchmark for evaluating long-form factuality in large language models. It comprises 2,280 fact-seeking prompts spanning 38 topics, designed to test a model's ability to generate accurate, long-form responses. The benchmark uses SAFE (Search-Augmented Factuality Evaluator) to evaluate factual accuracy.

  0.8%

  Unverified

• OpenAI-MRCR: 2 needle 256k

  #1 / 1

  Multi-Round Co-reference Resolution (MRCR) benchmark that tests long-context reasoning by evaluating a model's ability to distinguish between similar outputs, reason about ordering, and reproduce specific content from multi-turn conversations containing multiple writing requests on overlapping topics at 256k tokens.

  86.8%

  Unverified