Muse Wonder
An AI model released by Meta Superintelligence Labs and part of the Muse series. It is positioned as a flagship or standard-tier model intended to be competitive with leading models in the industry.
Core Facts
On the Artificial Analysis index, Claude Opus 5.5 scores 58, while Gemini 4 Argon and GPT-6 Astra both score 53.
Muse Wonder is offered through the Meta Model API and Muse Code; Muse Wonder Full is for Muse Code subscribers.
In the Intelligence Index given by the independent evaluation organization Artificial Analysis, Muse Wonder scores only 52, while Claude Opus 5.5 reaches 58.
Timeline (last 90 days)
Meta released two models, Muse Wonder and Muse Spark Open; Muse Wonder focuses on agents and professional work, and Muse Spark Open is an 816B-parameter MoE open-weights model.
The lava lamp test targeted the regular version of Muse Wonder, while some of the standout results Meta published came from Wonder Full, and there is a capability gap between the two.
Meta's official Model Card explicitly acknowledges that visual and spatial reasoning (especially 3D tasks) is a known weakness of Muse Wonder.
A single lava lamp failure only shows that the model did not complete the current task and cannot directly prove that it was over-optimized for public benchmark sets.
The developer community BridgeMind tested Muse Wonder with a 3D programming task (making a lava lamp), and the model produced only a red cylinder.
As of October 13, Meta has released only a 13-page model card, with the safety report 'coming soon'.
On Meta's own 15-test scorecard, Wonder took first place on 7.
Muse Wonder has a context window of about 1 million tokens, supports text and images as input, and outputs text only.
Meta released the AI model family Muse Wonder on October 12.
On the SWE Marathon test, MuseWonderFull scored 57.5, the regular version 53.1, Claude Opus 5.5 46.3, and GPT-6 Astra 42.5.
40 more timeline events
All Facts (20)
On the Artificial Analysis index, Claude Opus 5.5 scores 58, while Gemini 4 Argon and GPT-6 Astra both score 53.
80%VerifiedMuse Wonder is offered through the Meta Model API and Muse Code; Muse Wonder Full is for Muse Code subscribers.
80%VerifiedIn the Intelligence Index given by the independent evaluation organization Artificial Analysis, Muse Wonder scores only 52, while Claude Opus 5.5 reaches 58.
80%VerifiedMuse Wonder's API pricing is $4 per million input tokens and $20 per million output tokens, offered through Meta's coding service platform.
80%VerifiedMuse Wonder offers six reasoning levels: minimal, low, medium, high, xhigh, and max.
70%VerifiedThe Muse Wonder series includes two model variants: Wonder (a streamlined version optimized for speed) and WonderFull (the full version provided via API).
65%VerifiedMeta launched two new AI models: Muse Wonder and Muse Wonder Full.
75%VerifiedMeta released the frontier model Muse Wonder on October 12, and launched Wonder Full at the same event.
65%VerifiedMeta released the MuseWonder and WonderFull model series; Wonder is already available on the Meta model API, and WonderFull is used for MuseCoding.
65%Partially VerifiedMuse Spark Open can be called on the day of release through platforms such as Baseten, Fireworks AI, Groq, Modal, Nebius, OpenRouter and Together AI.
75%Partially VerifiedMuse Wonder supports a 1 million Token context window.
75%Partially VerifiedIn coding agent scenarios, Claude Code on Opus 5.5 scored 66, Wonder scored 63, and only Wonder Full can reach 68.
70%Partially VerifiedMuse Wonder offers three Meta API request formats (Responses, Chat Completions, Messages), Muse Code, and Vercel AI Gateway and Cloudflare Workers AI integrations.
70%Partially VerifiedMuse Wonder costs $4.52 per task call, an increase of nearly 3 times compared with Muse Spark 1.3, about $1 more expensive than GPT-6 Astra or Grok 4.7, but about $1.5 cheaper than Opus 5.5.
70%Partially VerifiedIn Artificial Analysis's calculated cost per task, Muse Wonder is $4.52 and Claude Opus 5.5 is $5.98.
65%Partially VerifiedMuse Wonder natively accepts text, image, video, and audio input, whereas Claude Opus 5.5 only supports text, image, and PDF.
65%Partially VerifiedIn the benchmark table published by Meta, Claude Opus 5.5 won 5 of 8 rows, while Muse Wonder won two software engineering metrics and a legal metric.
65%Partially VerifiedCached input still counts toward TPM; caching is a billing discount, not a rate-limit discount.
65%Partially VerifiedMuse Wonder supports a 1,048,576 Token context window (per request).
65%Partially VerifiedIn the DeepSWE 1.1 benchmark, Muse Wonder Full achieved 79.0%, exceeding Claude Opus 5.5's 74.2% and GPT-6 Astra's 74.1%.
65%