Command A+

Leaderboards

Where Command A+ 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 Command A+ has a published result on, out of 100%, with its rank among the models that took it.

• AIME 2025

  #50 / 124

  Competition maths problems from the 2025 American Invitational Mathematics Exam.

  90.0%

  Self-reported

• MMMU-Pro

  #51 / 74

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

  63.0%

  Self-reported · In the vision score

• MMMU

  #24 / 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.

  75.1%

  Self-reported

• CharXiv-R

  #56 / 61

  Reading and reasoning about charts from scientific papers.

  52.7%

  Self-reported

• IFBench

  #25 / 46

  Instruction Following Benchmark evaluating model's ability to follow complex instructions

  74.0%

  Self-reported

• MathVista

  #7 / 40

  MathVista evaluates mathematical reasoning of foundation models in visual contexts. It consists of 6,141 examples derived from 28 existing multimodal datasets and 3 newly created datasets (IQTest, FunctionQA, and PaperQA), combining challenges from diverse mathematical and visual tasks to assess models' ability to understand complex figures and perform rigorous reasoning.

  80.6%

  Self-reported

• Tau2 Telecom

  #20 / 37

  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.

  85.0%

  Self-reported

• SciCode

  #25 / 36

  SciCode is a research coding benchmark curated by scientists that challenges language models to code solutions for scientific problems. It contains 338 subproblems decomposed from 80 challenging main problems across 16 natural science sub-fields including mathematics, physics, chemistry, biology, and materials science. Problems require knowledge recall, reasoning, and code synthesis skills.

  38.0%

  Self-reported

• WMT24++

  #7 / 24

  WMT24++ is a comprehensive multilingual machine translation benchmark that expands the WMT24 dataset to cover 55 languages and dialects. It includes human-written references and post-edits across four domains (literary, news, social, and speech) to evaluate machine translation systems and large language models across diverse linguistic contexts.

  81.0%

  Self-reported

• CharXiv-D

  #8 / 18

  CharXiv-D is the descriptive questions subset of the CharXiv benchmark, designed to assess multimodal large language models' ability to extract basic information from scientific charts. It contains descriptive questions covering information extraction, enumeration, pattern recognition, and counting across 2,323 diverse charts from arXiv papers, all curated and verified by human experts.

  88.0%

  Self-reported

• Terminal-Bench Hard

  #2 / 2

  Terminal-Bench Hard is a harder terminal-agent benchmark variant evaluated with the Terminus-2 harness in Cohere's Command A+ and North Mini Code releases.

  25.0%

  Unverified

• Cohere Agentic Question Answering

  #1 / 1

  Cohere's internal North evaluation for measuring how well a model answers enterprise questions using MCP-connected cloud file systems. Scores are reported with LLM-as-a-judge techniques.

  65.0%

  Unverified

• Cohere Data Analysis

  #1 / 1

  Cohere's internal North evaluation for measuring a model's ability to perform data science tasks over uploaded spreadsheets. Scores are reported with LLM-as-a-judge techniques.

  45.0%

  Unverified

• Cohere Memory Usage Quality

  #1 / 1

  Cohere's internal North evaluation for measuring how well an agent uses information from North's memory system across sessions. Scores are reported with LLM-as-a-judge techniques.

  54.0%

  Unverified

• MT-AIME 2025

  #1 / 1

  MT-AIME 2025 is Cohere's internal multilingual translation of AIME 2025, evaluated for Arabic, Japanese, and Korean in the Command A+ release.

  86.0%

  Unverified