Meta's new AI models reach the top tier, but cost nearly three times more than before
Meta's Muse Wonder models score close to the best from OpenAI and Anthropic in tests. Running them costs nearly three times what Meta's older model did.
Meta released two new AI models on Monday, called Muse Wonder and Muse Wonder Full. It is Meta's first try at a model that competes with the best from OpenAI and Anthropic. Meta says its own tests show scores similar to those rivals' top models.
Outside scoring points the same way. Artificial Analysis, a firm that rates AI models, gave Wonder Full 56 points and Wonder 52. Meta's older model, Muse Spark 1.3, got 48. According to The Decoder, Wonder Full ranks in the top three on six of ten tests. The biggest gains are in coding and in doing tasks without a person guiding each step.
There is a weak spot. On a test of factual knowledge, Wonder Full scored 30. Anthropic's Claude Opus 5.5 scored 46. It does make things up less often than Meta's older model, falling from 33 to 23 percent.
The other catch is cost. Artificial Analysis says Wonder costs $4.52 per task, versus $1.60 for the older model. Wonder Full is only open to subscribers of Meta's coding product, Muse Code.
To be honest, most people won't notice anything yet. Meta says Wonder will power its Muse app, starting with paying subscribers, but it has given no date. For people who use AI to write code, Meta is now a real option, but at $4.52 a task it is not the cheapest one.
The facts
- Meta released Muse Wonder and Muse Wonder Full on Monday, its first attempt at a frontier-class model.
- According to Meta's own testing, the two models score comparably to Fable 5.1 and GPT-6 Astra.
- On the Artificial Analysis Intelligence Index, Muse Wonder Full scores 56 and Muse Wonder 52, up from 48 for Muse Spark 1.3.
- Muse Wonder Full places in the top three on six out of ten tests, according to The Decoder.
- Artificial Analysis puts Muse Wonder at $4.52 per task, nearly three times the $1.60 for Muse Spark 1.3.
- On the AA-Omniscience knowledge test, Muse Wonder Full scores 30 and Muse Wonder 25, versus 46 for Claude Opus 5.5.
- Muse Wonder Full cut its rate of made-up answers from 33 to 23 percent compared with its predecessor.
- Meta has not given specifics or a timeline for using Wonder in its Muse agent or for building specialized models for everyone.
Sources
- TechCrunch AI: Meta releases its first frontier model, Muse Wonder
- The Decoder: Meta's new frontier model Muse Wonder Full breaks into the AI top tier
- The Verge AI: Meta has some Wonder Full new AI models
Published
More in New models
DeepSeek says new downloadable AI beats pricier rivals at coding for about 20 cents a job
DeepSeek's own tests show a smaller model beating bigger rivals at coding, but outside tests are missing and China may limit such releases.
AmazonNew modelsAyobi’s pickReflection AI's Beam, a free-to-download coding AI, is now available on Amazon's cloud
Reflection AI released Beam, an AI built to write and fix software over long jobs. Anyone can download and change it, and Amazon now offers it to run on its cloud.
New modelsOfficial releaseAyobi’s pickvia Hugging FaceIBM releases AI models that chat, make quick calls and transcribe speech in one
IBM says its new Granite Switch 4.1 models handle conversation, yes-or-no style judgments and speech-to-text from one online connection, using the memory of one model.