Meta's Muse Wonder Beats Fable 5.1, Battles Sonnet 5.5 and Opus 5.5

Muse Wonder Full (max) showed strong gains across all ten Intelligence Index evaluations, with the largest improvements in agentic tasks. On Terminal-Bench 4.0, it jumped 24 points from 33% to 57%. On GDPval-AA v2.1, it added 176 Elo points, rising from 1681 to 1857. On AA-Briefcase, it gained 148 Elo, moving from 1581 to 1729.

The model places in the top three on six of the ten evaluations. It narrowly leads on GDP.pdf with 33%, ranks second on GDPval-AA v2.1, and third on both AA-Briefcase and AutomationBench-AA with 71%. In scientific reasoning, it tops the CritPt leaderboard at 33% and ranks third on SciCode with 63%.

• Terminal-Bench 4.0: 57% (+24 points)
• GDPval-AA v2.1: 1857 Elo (+176 points)
• AA-Briefcase: 1729 Elo (+148 points)
• GDP.pdf: 33% (leader)
• CritPt: 33% (leader)
• SciCode: 63% (third place)

Both new models use fewer output tokens than Muse Spark 1.3, consuming around 54,000 output tokens per Intelligence Index task, a decrease of about 10% from the previous model's 60,000. This is lower than similarly intelligent models like Gemini 4 Argon (high, 62k) and Claude Fable 5.1 (max w/ fallback, 78k), but higher than GPT-6.1 Sol (max, 38k) and GPT-6 Astra (max, 27k).

Muse Wonder Full (max) also uses 16% fewer input tokens per task than Muse Wonder (max), 6.44 million versus 7.69 million, making it both more intelligent and more token-efficient. Despite this efficiency, the premium API pricing and high input token use push Muse Wonder (max)'s cost to $4.52 per Intelligence Index task. This is lower than Claude Fable 5.1 ($7.63), Claude Opus 5.5 ($5.98), and Claude Sonnet 5.5 ($5.46), but higher than GPT-6 Astra ($3.26), Gemini 4 Argon ($1.99), GPT-6.1 Sol ($0.72), and Muse Spark 1.3 ($1.60). No cost per task estimate is available for Muse Wonder Full yet.

For users in Tunisia, Muse Wonder is accessible through Meta's first-party API and Muse Code. Muse Wonder Full is currently limited to Muse Code users, so wider availability may come later. Pricing for API access is set at $4 per million input tokens and $20 per million output tokens, which translates to roughly 11.6 TND and 58 TND respectively at current exchange rates.

These models are designed for personal and professional agent use, according to Meta. Developers and businesses looking for advanced agentic capabilities should consider Muse Wonder for tasks like automation and coding, as it outperforms Fable 5.1 on key benchmarks. There is no official announcement yet regarding specific availability or pricing in the Tunisian market beyond the API.