Models / muse-wonder

Muse Wonder

Meta's model, with closed weights. Forcebench has run it in 4 configurations on the API board. API is models reached through a hosted service; Local, models served on Forcebench’s own hardware.

On the API board

Models reached through a hosted service. Muse Wonder is ranked by the highest-scoring of its 4 configurations: the 12th-highest score of 29 models, not clearly different from 10 other models. Scores are within 3 attempts (c@3) with 95% intervals; each configuration is compared with the highest-scoring one by whether their 95% intervals overlap. Every configuration on the board → Cost and speed →

87%
within 3 attempts (c@3), 95% CI 84–91 · low · 12th-highest score of 29 models, not clearly different from 10 other models
2k
output tokens per task · median 1.6k, p90 3.5k
15 s
median per task, every attempt counted, including network and queueing · 90th percentile 52 s
0%
no answer within the 32,768-token budget
$0.061
per task at the list price of 20 October 2026, every attempt counted
$0.069
per solved task: the total cost over the tasks solved · cost and speed
Every configuration
Within 3 attempts (c@3) with 95% intervals, highest first. Bars whose intervals overlap are not clearly different.
  • 95% interval
low OpenRouterhighest score
87%, 95% interval 84–91%
medium OpenRouter
87%, 95% interval 84–91%
max OpenRouter
87%, 95% interval 84–91%
high OpenRouter
86%, 95% interval 82–89%
    Show the numbers
    Every configuration of Muse Wonder on the API board
    ConfigurationRank all configsWithin 3 attempts c@3 · 95% CIAgainst the highest by 95% interval overlapTokens mean · median · p90No answerLatency median per task, every attempt countedCost per taskCost per solved task
    OpenRouter · lowhighest score19
    87%95% CI 84–91
    		—2k1.6k · 3.5k0%15 s$0.061$0.069
    OpenRouter · medium19
    87%95% CI 84–91
    		not clearly different3.1k2.4k · 5.9k0%22 s$0.084$0.096
    OpenRouter · max19
    87%95% CI 84–91
    		not clearly different8.2k5.8k · 20k1%55 s$0.22$0.25
    OpenRouter · high23
    86%95% CI 82–89
    		not clearly different4.4k3.1k · 9.2k0%33 s$0.13$0.15
    Suites, at its highest-scoring configuration
    low · OpenRouter: within 3 attempts (c@3) on each suite with 95% intervals, against the median of the API board's models, each at its highest-scoring configuration.
    • 95% interval
    • API board median
    Apex
    95%, 95% interval 76–99%
    Salesforce APIs
    83%, 95% interval 61–94%
    CI/CD
    88%, 95% interval 66–97%
    Salesforce CLI
    100%, 95% interval 84–100%
    Salesforce docs
    35%, 95% interval 17–59%
    fflib Enterprise Patterns
    89%, 95% interval 67–97%
    Flow
    83%, 95% interval 61–94%
    Governor limits & pushback
    100%, 95% interval 83–100%
    Lightning Web Components
    100%, 95% interval 82–100%
    Nonprofit Success Pack
    100%, 95% interval 82–100%
    Packaging
    89%, 95% interval 67–97%
    Permissions & access
    93%, 95% interval 70–99%
    Scratch org definitions
    67%, 95% interval 44–84%
    SOQL
    95%, 95% interval 76–99%
    Trigger Actions Framework
    94%, 95% interval 74–99%
      Show the numbers
      Per-suite scores of Muse Wonder's highest-scoring configuration on the API board, with the board's median
      Suitec@3 95% CIAPI medianTasks
      Apex
      95%95% CI 76–99
      		90%20
      Salesforce APIs
      83%95% CI 61–94
      		72%18
      CI/CD
      88%95% CI 66–97
      		88%17
      Salesforce CLI
      100%95% CI 84–100
      		90%20
      Salesforce docs
      35%95% CI 17–59
      		29%17
      fflib Enterprise Patterns
      89%95% CI 67–97
      		61%18
      Flow
      83%95% CI 61–94
      		78%18
      Governor limits & pushback
      100%95% CI 83–100
      		89%19
      Lightning Web Components
      100%95% CI 82–100
      		100%18
      Nonprofit Success Pack
      100%95% CI 82–100
      		83%18
      Packaging
      89%95% CI 67–97
      		72%18
      Permissions & access
      93%95% CI 70–99
      		87%15
      Scratch org definitions
      67%95% CI 44–84
      		50%18
      SOQL
      95%95% CI 76–99
      		100%20
      Trigger Actions Framework
      94%95% CI 74–99
      		89%18

      By reasoning effort

      The same configuration at several efforts, each against its lowest reasoning effort on the same tasks (a paired 95% interval), and charted against the effort tier and against the output tokens it costs. These scores are pass@1, the first answer alone: a paired comparison needs both efforts' answers to the same tasks, and not every effort has had its retries. Lite-subset sweeps (4 tasks per suite) are never plotted with the full benchmark. The effort study has every model.

      Muse Wonder by reasoning effort, full benchmark: each effort against the sweep's lowest reasoning effort
      Configuration · effort · full benchmarkpass@1 95% CIAgainst the lowest reasoning effort same tasks, pairedTokens
      lowbaseline
      70%95% CI 65–75
      		—2k
      medium
      77%95% CI 73–82
      		clearly higher (+7 pts, 95% CI +3 pts to +12 pts)3.1k
      high
      69%95% CI 64–74
      		no clear difference (−1 pt, 95% CI −5 pts to +4 pts); this data can't rule out a change of up to 5 points4.4k
      max
      74%95% CI 69–78
      		no clear difference (+4 pts, 95% CI −0.7 pts to +8 pts); this data can't rule out a change of up to 8 points8.2k
      Score by effort tier · full benchmark
      Overall pass@1; vertical bars are 95% confidence intervals.
      Muse Wonder, full benchmark: pass@1 by reasoning effort tierMuse Wonder: low 70%, medium 77%, high 69%, max 74%0%25%50%75%100%LowMediumHighMaxReasoning effort tierpass@1Muse Wonder
      Show the numbers
      ModelEffort (tier)pass@195% CITasksNo answerTokens: meanmedianReasoning tokensLatency: median per answer
      Muse Wonderlow70%65–752720%2k1.6k1.4k15 s
      Muse Wondermedium77%73–822720%3.1k2.4k2.5k22 s
      Muse Wonderhigh69%64–742720%4.4k3.1k3.8k30 s
      Muse Wondermax74%69–782721%8.2k5.8k7.6k53 s
      Score against output tokens · full benchmark
      Each point is one effort setting. Further right means more output tokens per task. The numbers are under the chart above.
      Muse Wonder, full benchmark: pass@1 against mean output tokens per taskMuse Wonder: 2k tokens 70%, 3.1k tokens 77%, 4.4k tokens 69%, 8.2k tokens 74%0%25%50%75%100%2k5k10kMean output tokens per task (log scale)pass@1lowmediumhighmaxMuse Wonder

      In the studies

      Private evaluations

      Want this for your team?

      Azul Labs runs Forcebench privately for two kinds of Salesforce team.

      Choosing a model to run yourself?

      A private benchmark of your shortlist

      Open-weight models at the quantisations and reasoning efforts you could run on infrastructure you control, tested on your kind of Salesforce work, with a written recommendation.

      • Private tasks written from your work only ever run on local models; they are never sent to an API.
      • Re-testing after any change of model, quantisation or engine is available.
      • Typical turnaround is 2–3 weeks; large models take about a day of machine time per configuration.
      Get a private benchmarkHow it works

      Building or buying a Salesforce AI tool?

      A private evaluation of your tool

      Your tool on held-out tasks it has never seen, against API baselines (models reached through a hosted service) and Local ones (served on Forcebench’s own hardware), with a written report on where it holds up and where it fails.

      • It can start as a smaller pilot, about 30 held-out tasks with your tool against two baselines, before a full evaluation.
      • Repeat evaluations after your releases are available.
      • Typical turnaround is 2–3 weeks.
      Ask about an evaluationHow it works

      Fixed fee, quoted up front. The quote depends on the number of tasks and the number of models, tools or configurations tested.

      Or email leo@azl.au · azl.au