Kimi K3.1 Pro

Leaderboards

Where Kimi K3.1 Pro 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 Kimi K3.1 Pro has a published result on, out of 100%, with its rank among the models that took it.

• GPQA

  #5 / 265

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

  94.2%

  Self-reported

• Humanity's Last Exam

  #22 / 115

  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

  51.1%

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

• MMMU-Pro

  #2 / 74

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

  84.1%

  Self-reported · In the vision score

• BrowseComp

  #1 / 73

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

  92.1%

  Self-reported · In the agents score

• CharXiv-R

  #10 / 61

  Reading and reasoning about charts from scientific papers.

  89.0%

  Self-reported

• Terminal-Bench 4.0

  #3 / 31

  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.

  65.7%

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

• Agents' Last Exam

  #11 / 26

  Agents' Last Exam is a challenging benchmark for AI agents on hard, long-horizon tasks that test sustained reasoning, planning, and tool use, reported with and without tool access.

  43.2%

  Self-reported · In the agents score

• BabyVision

  #1 / 18

  A benchmark for early-stage visual reasoning and perception on child-like vision tasks.

  95.4%

  Self-reported

• ZEROBench

  #6 / 16

  ZEROBench is a challenging vision benchmark designed to test models on zero-shot visual understanding tasks.

  0.4%

  Self-reported

• GDP.pdf

  #3 / 15

  GDP.pdf is a knowledge-work vision benchmark that evaluates models on economically valuable professional tasks presented as visual documents (PDFs), testing document-based reasoning, chart and table interpretation, and problem solving without tools.

  34.6%

  Self-reported

• DeepSWE

  #1 / 14

  DeepSWE is a software engineering agent benchmark evaluated with the mini-swe-agent harness, where each task is solved in an isolated container with no internet access. It measures an agent's ability to autonomously resolve real-world coding issues end to end.

  75.5%

  Self-reported

• APEX-Agents

  #1 / 13

  APEX-Agents is a benchmark evaluating AI agents on long horizon professional tasks that require sustained reasoning, planning, and execution across complex multi-step workflows.

  67.2%

  Self-reported

• OfficeQA Pro

  #4 / 13

  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.

  67.4%

  Self-reported

• SWE-Marathon

  #5 / 11

  SWE-Marathon is an ultra-long-horizon software engineering benchmark covering tasks such as building compilers, optimizing kernels, and developing production-grade services. It measures whether agents can sustain quality across extremely long engineering trajectories.

  51.7%

  Self-reported

• ExploitGym

  #5 / 10

  ExploitGym is a large-scale, realistic benchmark built from real-world vulnerabilities across userspace programs, Google's V8 engine, and the Linux kernel. Given a vulnerability and a proof-of-vulnerability input, agents must craft a working end-to-end exploit that achieves unauthorized code execution.

  20.7%

  Self-reported

• PostTrainBench

  #6 / 10

  PostTrainBench evaluates a model's ability to autonomously post-train base models. Given pretrain-only base models, the agent must complete the full pipeline of data synthesis, training, evaluation, and iteration within a time budget, scored across downstream benchmarks such as AIME2025, BFCL, GPQA Main, GSM8K, and HumanEval.

  36.5%

  Self-reported

• Terminal-Bench-Science 0.1

  #9 / 10

  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.

  38.1%

  Self-reported

• ExploitBench

  #3 / 9

  ExploitBench is a cybersecurity benchmark that evaluates a model's ability to discover and exploit software vulnerabilities, reported as the fraction of challenges where the model captures the target (Cap%).

  74.7%

  Self-reported

• BenchCAD

  #3 / 8

  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.

  72.8%

  Self-reported

• BenchCAD (with Python tool)

  #4 / 8

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

  92.7%

  Self-reported

• BioMysteryBench

  #4 / 8

  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.

  80.6%

  Self-reported

• Chartography

  #1 / 7

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

  90.2%

  Self-reported

• AA-Briefcase v1.1

  #3 / 6

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

  55.9%

  Self-reported

• AutomationBench-AA

  #1 / 6

  AutomationBench-AA is Artificial Analysis's independently run version of AutomationBench, covering 657 real-world SaaS workflow tasks across 40 simulated applications (e.g. Gmail, Slack, Salesforce, HubSpot). It scores the share of objectives an agent completes without violating business guardrails, using a private held-out task set.

  73.3%

  Self-reported

• GDPval-AA 2.1

  #1 / 6

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

  61.5%

  Self-reported

• Humanity's Last Exam (with tools)

  #1 / 6

  Full multimodal Humanity's Last Exam evaluated with tool use enabled, distinct from its text-only subset.

  70.0%

  Self-reported

• Harvey's Legal Agent Benchmark

  #1 / 5

  Harvey's Legal Agent Benchmark evaluates AI agents on complex legal workflows and reports the all-pass rate.

  23.7%

  Self-reported

• MMMU-Pro (with tools)

  #1 / 5

  MMMU-Pro variant evaluated with tool access enabled.

  90.0%

  Self-reported

• Kimi Code Bench v2

  #1 / 3

  Kimi Code Bench v2 is Moonshot AI's in-house benchmark for evaluating coding agents on realistic software engineering tasks across 10+ mainstream programming languages and a production tech stack spanning backend services, infrastructure, performance engineering, systems programming, security, frontend development, and ML/data engineering.

  77.1%

  Self-reported

• PerceptionBench

  #1 / 3

  PerceptionBench is Moonshot AI's internal benchmark for evaluating atomic visual perception capabilities.

  68.3%

  Self-reported