Muse Wonder Full (max) outperforms Fable 5.1 (max) on Artificial Analysis' Intelligence Index, matches Sonnet 5.5 (max), and is just shy of Opus 5.5 (max)
@artificialanlys:
@artificialanlys
Context & Ripple Effects
Meta’s Muse line had already climbed from Spark 1.2’s score of 54 in August to Spark 1.3’s score of 62 in a limited partner preview in September, when it still trailed Fable 5.1 and Opus 5. That makes Wonder Full’s placement an unusually rapid move into the index’s leading group.
The comparison also lands amid a growing efficiency question in frontier rankings: Sonnet 5.5 ranked near the top of the index but used the most tokens among the models compared. Artificial Analysis separately says Muse Wonder’s knowledge performance is weaker than its broader results, though its hallucination rates are low.
First-order effects
Artificial Analysis’ ranking moves Meta’s Muse family ahead of Fable 5.1 and level with Sonnet 5.5 on the cited configuration, narrowing the perceived capability gap among the leading proprietary models.
For buyers assessing max-reasoning configurations, Artificial Analysis puts Muse Wonder’s index-task cost at $4.52 versus $1.60 for Muse Spark 1.3, making the performance gain a direct cost-per-task trade-off.
Second-order effects
Fable 5.1 and Sonnet 5.5 must compete not only on composite rank but on the efficiency and evaluation-specific results that procurement teams can use to distinguish closely ranked models.
Muse Wonder’s split between broad benchmark strength and weaker knowledge scores gives enterprise evaluators reason to test factual-knowledge workloads separately from low-hallucination performance rather than rely on the aggregate index alone.
Third-order effects
If leading-model gains continue to require more tokens and higher task costs, frontier competition will be judged increasingly by performance per useful task rather than leaderboard position alone.
The variation across Artificial Analysis’ component evaluations points toward benchmark portfolios—covering reasoning, knowledge, reliability and cost—becoming more influential than any single composite score.
The trend: Frontier AI procurement is shifting from a race for the highest composite score toward multi-axis comparisons of capability, reliability and inference cost.
Knowledge remains a relative weakness for both models, but this is mitigated by very low hallucination rates. On AA-Omniscience, Muse Wonder Full (max) scores 30 and Muse Wonder (max) scores 25, level with Muse Spark 1.3 (max), below e.g. Claude Opus 5.5 (46), GPT-6 Astra (43), G…
Muse Wonder Full (max) places in the top three on six of the ten Artificial Analysis Intelligence Index v4.3 evaluations: ➤ #1 on CritPt (33%) and GDP.pdf (33%) ➤ #2 on GDPval-AA v2.1 (1857 Elo) ➤ #3 on AA-Briefcase (1729 Elo), AutomationBench-AA (71%) and SciCode (63%) The leads…
Muse Wonder (max) costs $4.52 per Artificial Analysis Intelligence Index task, 2.8x Muse Spark 1.3 (max, $1.60), driven by higher per-token prices and higher input token use ➤ Token pricing: $4 per 1M input tokens, $0.20 cached and $20 output, up from $1.25 / $0.15 / $4.25 for Mu…
Is Muse Wonder full better than opus 5.5? I gave them the same prompt goal is to build me a triple aaa level lawn mower simulator i can drink beer and smoke a cigar. Api equivalent cost : muse - $4.11 opus - $99.43 It is very clear that muse is not anywhere near close to frontier…
Try very hard to avoid commenting on releases but as a general comment, I think publishing intelligence without cost is not-great and I hope it doesn't become a trend [embedded post]
1. @Meta is now in the frontier lab convo with their “Muse Wonder” model 2. Despite @OpenAI seemingly leapfrogged, this is a marathon not a sprint. 3. wtf @Google. What's wrong with y'all? [embedded post]