Muse Wonder Full

Leaderboards

Where Muse Wonder Full 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 Muse Wonder Full has a published result on, out of 100%, with its rank among the models that took it.

• DeepSWE 1.1

  #1 / 50

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

  79.0%

  Self-reported · In the coding score

• HealthBench Professional

  #7 / 19

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

  62.3%

  Self-reported

• Program Bench

  #6 / 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.

  59.0%

  Self-reported · In the coding score

• GDP.pdf

  #4 / 14

  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.

  33.2%

  Self-reported

• Artificial Analysis

  #3 / 13

  Artificial Analysis benchmark evaluates AI models across quality, speed, and pricing dimensions, providing a composite assessment of model capabilities for real-world usage.

  56.0%

  Self-reported

• SWE-Marathon

  #1 / 10

  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.

  57.5%

  Self-reported

• JobBench

  #1 / 6

  JobBench evaluates agents on work-related tasks, as reported in the MiMo-V2.6 release.

  75.3%

  Self-reported

• OSWorld 2.1 (partial)

  #2 / 6

  Partial-credit computer-use score on OSWorld 2.1, as reported in the Claude Sonnet 5.5 release. Separate from OSWorld 2.0 and strict completion scores.

  78.4%

  Self-reported

• OSWorld 2.1 (strict)

  #2 / 6

  Strict pass rate on OSWorld 2.1: fraction of tasks where every graded checkpoint is satisfied. Separate from the OSWorld 2.1 partial-credit score and from OSWorld 2.0.

  39.4%

  Self-reported

• Vals Finance Agent v2

  #3 / 6

  Vals Finance Agent v2 evaluates AI agents on realistic financial analyst tasks.

  59.1%

  Self-reported

• AutomationBench-AA

  #2 / 5

  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.

  45.4%

  Self-reported

• Harvey LAB-AA

  #2 / 5

  Harvey LAB-AA evaluates model performance on complex legal workflows.

  13.1%

  Self-reported

• FocusBench

  #1 / 4

  Internal personal agent evaluation across housing, travel, IT, and finance; persistence through blockers and adaptation across interactions and changing environments.

  54.2%

  Self-reported

• MemoryBench

  #3 / 4

  Internal recall evaluation over 64K-100K-token conversations containing distractors, conflicting facts, and later updates; LLM judge with weighted question-specific rubrics.

  85.5%

  Self-reported

• Personal Superintelligence Index

  #1 / 4

  Internal composite of Agents' Last Exam, AutomationBench-AA, FocusBench, JobBench, GDP.pdf, VBT, and DRACO.

  59.1%

  Self-reported

• Terminal-Bench 4.0-AA

  #1 / 4

  Artificial Analysis reported accuracy on terminal tasks with executable final-container-state verifiers.

  57.1%

  Self-reported