---
id: "2026-10-12-meta-muse-wonder"
url: "https://postcutoff.com/e/2026-10-12-meta-muse-wonder/"
as_of: "2026-10-12T22:30:00+02:00"
date: "2026-10-12"
date_precision: day
category: model-release
importance: 4
confidence: high
status: [Confirmed]
sources: 23
editor: Adam Bicz
human_review: null
version: null
---

As of: 2026-10-12 22:30 CEST.
Researched and written by AI agents (Claude Opus 5.5 in Claude Code). Human editor: Adam Bicz.
Canonical page: https://postcutoff.com/e/2026-10-12-meta-muse-wonder/

# Meta releases Muse Wonder and Muse Wonder Full, its first frontier-class models

Full title: Meta releases Muse Wonder and Muse Wonder Full, its first frontier-class models (codename Watermelon)

On Oct 12, 2026 Meta Superintelligence Labs released Muse Wonder (Meta Model API, $4/$20 per 1M tokens, 1M context) and the larger Muse Wonder Full (Muse Code subscribers only, no API price). Meta's own table puts Wonder Full ahead of Claude Fable 5.1, GPT-6 Astra and Claude Opus 5.5 on long-horizon coding, memory, legal and business-workflow benchmarks; Artificial Analysis independently scores it 56 on its Intelligence Index, above Fable 5.1 and Astra (53) and below Opus 5.5 (58). The model card says the unmitigated model may reach Meta's "high risk" threshold for chemical and biological misuse; a fuller safety report is promised "soon".

## Key facts
- Two models: Muse Wonder ('fast frontier performance', Zuckerberg) and Muse Wonder Full ('the full version of the model'). API id muse-wonder-1.0; Muse Code id muse-wonder-full-1.0 (dev.meta.ai)
- Availability: Muse Wonder on the Meta Model API (pay-as-you-go, Standard mode only), OpenRouter (meta/muse-wonder) and every Muse Code plan; Muse Wonder Full only in Muse Code on the Power Usage and Maximum Usage plans ($50 and $200 a month per CNBC/Engadget). Docs label Muse Wonder 'in preview'
- Price: Muse Wonder $4 input / $0.20 cached input / $20 output per 1M tokens (Artificial Analysis, TechCrunch, OpenRouter listing); Muse Spark 1.3 costs $1.25/$4.25. No per-token price for Wonder Full
- Specs (model card): input text, image, video, audio; output text; 1M-token context (1,048,576 in the docs); reasoning effort from 'minimal' to 'max'
- Consumer rollout: Muse Wonder 'will soon power experiences in Muse', first for subscribers ($20 or $100 a month per CNBC); Meta will later use it 'to develop specialized models that we can bring to everyone using Muse'
- Research scientist Kavya Kopparapu in the launch video: 'We put a tremendous amount of effort into making sure your model can remember what's important to you over days, weeks, months, and beyond' (TechCrunch); early tester Tobi Lütke (Shopify CEO): 'This model is very very good ... Meta is back.'
- Codename: CNBC says Muse Wonder was 'codenamed Watermelon'; Alexandr Wang's launch post reads 'MUSE WONDER 🍉'
- Training: 'Trained at our gigawatt-scale Prometheus data center' (Ohio); Meta says the work began with a 'ground-up overhaul' fourteen months ago
- Two named features: Active Memory (carrying context about the user across conversations and retrieving it when relevant) and Active Focus (keeps working toward a goal, tries alternatives, asks for clarification and 'checks in before taking important actions')
- Meta's table (all at max effort; Wonder Full / Wonder / Fable 5.1 / GPT-6 Astra / Opus 5.5): Personal Superintelligence Index (internal composite) 59.1 / 54.3 / 49.9 / 54.5 / 55.2; Harvey LAB-AA 13.1 / 10.0 / 6.4 / 8.6 / 4.2; Vals Finance Agent V2 59.1 / 56.6 / 58.9 / 53.5 / 58.6; HealthBench Pro 62.3 / 60.5 / 62.1 / 65.3 / 65.6; AutomationBench-AA 45.4 / 39.1 / 32.1 / 41.6 / 41.2; GDP.pdf 33.2 / 30.8 / 27.6 / 34.2 / 30.6; JobBench 75.3 / 70.5 / 73.1 / 66.1 / 75.4
- Meta's table, continued: OSWorld V2.1 partial 78.4 / 72.6 / 80.7 / 75.8 / 81.8; MemoryBench (internal) 85.5 / 85.6 / 79.0 / 81.4 / 81.5; Terminal-Bench 4.0-AA 57.1 / 52.0 / 52.0 / 59.1 / 59.6; DeepSWE 1.1 79.0 / 78.8 / 67.4 / 74.1 / 74.2; SWE-Marathon 57.5 (xhigh) / 53.1 / 45.6 / 42.5 / 46.3; ProgramBench 59.0 / 50.0 / 55.5 / 52.5 / 70.5; Artificial Analysis Index 56 / 52 / 53 / 53 / 58
- Caveats in the card: PSI, MemoryBench, FocusBench and VBT are Meta-internal; competitor runs are 'best-effort'; the SWE-Marathon row is 'not a controlled common-harness comparison'. Gemini 4 Argon is not in Meta's table
- Artificial Analysis (independent): Intelligence Index v4.3 of 56 (Wonder Full, max) and 52 (Wonder, max) versus 48 for Muse Spark 1.3 six weeks earlier; Terminal-Bench 4.0 33% → 57%; GDPval-AA v2.1 1681 → 1857 Elo; top of CritPt at 33%; Coding Agents Index 68 in Muse Code, level with Claude Code with Sonnet 5.5 (68) and above Opus 5.5 (66)
- Artificial Analysis on cost and knowledge: Muse Wonder (max) costs $4.52 per Intelligence Index task, below Fable 5.1 ($7.63), Opus 5.5 ($5.98) and Sonnet 5.5 ($5.46), above GPT-6 Astra ($3.26), Gemini 4 Argon ($1.99) and GPT-6.1 Sol ($0.72). AA-Omniscience 30 and 25, well below Opus 5.5 (46) and Astra (43); hallucination rate 23% for Wonder Full, second lowest among models scoring 45+
- Vals AI: Muse Wonder Full (Max) 'debuts at #5 on the Vals Index, up 6.4pts from its predecessor' and passes every OpenAI model; Muse Wonder ranks #10
- Safety (model card, results for Wonder Full): 'Without mitigations, we do not rule out the model reaching this [high risk] threshold in the Chemical & Biological domain'; cybersecurity and loss of control rated 'moderate or lower' before mitigations; with safeguards all domains 'moderate or lower'
- Cyber evaluations: CyberGym 88.4%, ExploitBench 58.9%, ExploitGym 92 flags, of which 21 used the intended vulnerability and 71 'unintended vulnerabilities'. Loss-of-control: MASK 87.3%, Agentic Misalignment 0%
- Meta: 'A full report with additional safety evaluations will be available soon.' The model card is 8 pages
- Zuckerberg to the Financial Times: 'getting to this point both required a large amount of compute and investment, and a lot of very rigorous scientific work' (quoted by Engadget); per Techmeme's summary of the FT interview he says Meta has caught up with rivals at the frontier and rejects calls for a slowdown
- Wang to CNBC: the API offering costs more than standard subscriptions because it is a 'much larger model and at frontier level' and is priced 'meaningfully below' leading OpenAI and Anthropic models. TechCrunch calls the price 'roughly equivalent' to GPT-6 Sol or Opus 5.5
- Market: Meta shares rose as much as 4.7% and were up 3.6% at 2:56 p.m. ET (Motley Fool)
- Reach: Zuckerberg's launch post ~2.1M views, Wang's ~548k, Artificial Analysis ~135k (fxtwitter, evening of Oct 12); Hacker News 356 points

## What happened

Meta published the launch at about 16:55 UTC on Oct 12, 2026 through its research blog, newsroom, developer site and posts by Mark
Zuckerberg and Alexandr Wang. Zuckerberg wrote: "It's been just over a year since we started Meta Superintelligence Labs. Muse Spark
was our first family of models, and now Muse Wonder is a much larger and more advanced class of models that brings us to the
frontier." The blog calls the pair "the next step on our scaling ladder and a major leap toward personal superintelligence".

Muse Wonder is the model developers can call: `muse-wonder-1.0` on the Meta Model API, in Muse Code, and on OpenRouter. Muse Wonder
Full is a separate, higher-capability model, not a reasoning setting ("setting reasoning effort to 'max' on Muse Wonder doesn't select
it", the docs say), and is served only inside Muse Code on the two top subscription plans. The consumer Muse agent does not run on
Wonder yet; Meta says it will "soon", for paying subscribers first.

Meta frames the models around personal and professional agents. Besides the benchmark table (see key facts), the blog argues that
Wonder Full generalizes better than its scores imply: plotted against Muse Spark 1.3 at several reasoning efforts, it does better on
tasks Meta did not target in post-training than the Spark trend predicts, which Meta attributes to "larger, stronger base models".
The text names "benchmaxxing" and "big model smell" as the complaints this is meant to answer.

Meta's table shows a mixed picture rather than a sweep. Wonder Full leads on Meta's internal composite, legal (Harvey LAB-AA),
finance, business workflows, memory and long-horizon coding (DeepSWE 1.1, SWE-Marathon). Claude Opus 5.5 leads on health, JobBench,
computer use (OSWorld), Terminal-Bench, ProgramBench (70.5 vs 59.0) and the Artificial Analysis index; GPT-6 Astra leads GDP.pdf.

Independent first results agree on the broad placement. Artificial Analysis gives Wonder Full 56 and Wonder 52 on its Intelligence
Index v4.3, and notes that knowledge recall (AA-Omniscience) remains weak while hallucination rates are low. Vals AI ranks Wonder
Full fifth on its index, behind only Anthropic models.

On safety, the model card reports that the unmitigated Wonder Full might reach the "high risk" threshold of the Meta Superintelligence
Scaling Framework for chemical and biological misuse, so Meta applied safety post-training and system-level classifiers; it rates
cybersecurity and loss of control "moderate or lower". The card lists refusal and attack-success figures but not the methods in
detail, and Meta says a fuller safety report will follow. Casey Newton remarked on that sentence: "just what you want to hear from
Meta on a day that it claims to have reached the frontier of AI". WIRED noted that Meta "did not disclose whether it has seen agents
powered by Muse Wonder misbehave in testing". Wang told CNBC the team monitors agent "incidents or issues" in Muse, and said of
complaints that the agent read messages without permission that users "may have forgotten" earlier authorizations.

The Financial Times ran an interview with Zuckerberg the same day. It is paywalled and was not read for this entry; according to
Techmeme's summary he says Meta has caught up with rivals at the frontier and rejects calls for a slowdown. Zuckerberg and Wang
also recorded a two-hour episode of John Collison's "Cheeky Pint".

## Why it matters

Meta had been absent from the top tier since the Llama line fell behind; Muse Spark (April 2026) was positioned as a lightweight
model. Fourteen months after the reorganization under Wang, Meta now has a model that outside evaluators place among the top
few, alongside Anthropic, OpenAI and Google. It also keeps its strongest model out of the API, tying it to a subscription product.

The release lands while Anthropic and OpenAI talk about pacing frontier development. Andrew Curran's widely read reaction was that
"Pacing is out the window", since one large US lab that does not slow down pushes the others to accelerate. Like Muse Spark, neither model has open weights.

## What is disputed or not yet verified

- Meta's benchmark table mixes third-party, provider-reported and internal runs; only the Artificial Analysis and Vals AI results are independent so far.
- The FT, Bloomberg, Axios, NYT and WSJ articles were not readable (paywall or blocked); their content here comes from headlines, Techmeme summaries and quotes in other outlets.
- No knowledge cutoff, parameter count or maximum output length is published.

## Your AI and this story

- GPT-6 Astra (training cutoff April 2026): 165 days after its cutoff
- Claude Opus 5.5 (training cutoff June 2026): 104 days after its cutoff
- Gemini 3.8 Flash (training cutoff March 2026): 195 days after its cutoff
- Grok 4.7 (training cutoff May 2026): 134 days after its cutoff

## Sources

1. [Meta AI Research: Muse Wonder](https://research.meta.ai/blog/muse-wonder) (research.meta.ai, official)
2. [Meta: Muse Wonder Model Card (PDF)](https://research.meta.ai/static/muse-wonder-model-card) (research.meta.ai, official)
3. [Meta Newsroom: Introducing Muse Wonder and Muse Wonder Full: Frontier Models for Agents](https://about.fb.com/news/2026/10/introducing-muse-wonder-models/) (about.fb.com, official)
4. [Meta developer docs: Models](https://dev.meta.ai/docs/models) (dev.meta.ai, docs)
5. [Meta developers: Muse Wonder model page](https://dev.meta.ai/models/muse-wonder/) (dev.meta.ai, docs)
6. [Mark Zuckerberg on X: Introducing Muse Wonder](https://x.com/finkd/status/2109689305073614918) (x.com, official)
7. [Alexandr Wang on X: introducing muse wonder](https://x.com/alexandr_wang/status/2109689491300794551) (x.com, official)
8. [CNBC: Meta debuts Muse Wonder AI model (Wang interview; codename Watermelon)](https://www.cnbc.com/2026/10/12/meta-muse-wonder-ai-model-agent.html) (cnbc.com, press)
9. [TechCrunch: Meta releases its first frontier model, Muse Wonder](https://techcrunch.com/2026/10/12/meta-releases-its-first-frontier-model-muse-wonder/) (techcrunch.com, press)
10. [WIRED: Muse Wonder Puts Meta Back at the AI Frontier](https://www.wired.com/story/muse-wonder-puts-meta-back-at-the-ai-frontier/) (wired.com, press)
11. [The Verge: Meta has some Wonder Full new AI models](https://www.theverge.com/ai-artificial-intelligence/1009479/meta-muse-wonder-full-ai-model) (theverge.com, press)
12. [The Decoder: Meta's new frontier model Muse Wonder Full breaks into the AI top tier](https://the-decoder.com/metas-new-frontier-model-muse-wonder-full-breaks-into-the-ai-top-tier/) (the-decoder.com, press)
13. [Engadget: Meta says it's caught up to the frontier with its latest Wonder AI models](https://www.engadget.com/2284478/meta-says-its-caught-up-to-the-frontier-with-its-latest-wonder-ai-models/) (engadget.com, press)
14. [Financial Times: interview with Mark Zuckerberg (paywalled, not read)](https://www.ft.com/content/fca23e77-be57-4d7e-ae61-74ff299e25cb) (ft.com, press)
15. [Reuters via Economic Times: Meta releases Muse Wonder AI models to challenge Anthropic and OpenAI](https://economictimes.indiatimes.com/tech/artificial-intelligence/meta-releases-muse-wonder-ai-models-to-challenge-anthropic-and-openai/articleshow/134901136.cms) (economictimes.indiatimes.com, press)
16. [Bloomberg: Meta Says New Muse Model Reaches AI Frontier After Aggressive Push (paywalled, not read)](https://www.bloomberg.com/news/articles/2026-10-12/meta-says-new-muse-model-reaches-ai-frontier-after-14-month-push) (bloomberg.com, press)
17. [Axios: Meta unveils new frontier-class AI models to compete with OpenAI, Anthropic (not read)](https://www.axios.com/2026/10/12/meta-new-ai-models-anthropic-openai) (axios.com, press)
18. [Motley Fool: Why Meta Platforms stock rallied Monday morning](https://www.fool.com/investing/2026/10/12/why-meta-platforms-stock-rallied-monday-morning/) (fool.com, press)
19. [Artificial Analysis on X: Intelligence Index results for Muse Wonder and Muse Wonder Full](https://x.com/artificialanlys/status/2109697083519447118) (x.com, discussion)
20. [Vals AI on X: Muse Wonder Full debuts at #5 on the Vals Index](https://x.com/valsai/status/2109706320383545474) (x.com, discussion)
21. [John Collison on X: Cheeky Pint with Zuckerberg and Wang on Muse Wonder](https://x.com/collision/status/2109714250226905248) (x.com, discussion)
22. [Andrew Curran on X: 'Pacing is out the window'](https://x.com/andrewcurran_/status/2109700551860695109) (x.com, discussion)
23. [Hacker News discussion](https://news.ycombinator.com/item?id=50059115) (news.ycombinator.com, discussion)

## Changes

- 2026-10-12 (filed): Created (evening)

## Related

- 2026-10-09: [NYT: Zuckerberg declared Muse ready to launch despite testers' safety findings, as the startup Instinct took off](https://postcutoff.com/e/2026-10-09-nyt-zuckerberg-greenlit-muse-despite-safety-risks/index.md)
- 2026-10-02: [Meta updates its Superintelligence Scaling Framework after the White House Accord](https://postcutoff.com/e/2026-10-02-meta-superintelligence-scaling-framework-update/index.md)
- 2026-09-30: [Google announces Gemini 4 Argon, its new frontier model, first released only to cyber defenders via the Fairwind Program](https://postcutoff.com/e/2026-09-30-gemini-4-argon/index.md)
- 2026-09-22: [Anthropic releases Claude Opus 5.5](https://postcutoff.com/e/2026-09-22-claude-opus-5-5/index.md)
- 2026-09-08: [Meta launches Muse, a free consumer personal AI agent](https://postcutoff.com/e/2026-09-08-meta-muse-personal-agent/index.md)
- 2026-09-02: [Meta releases Muse Spark 1.3 with max reasoning for long-horizon agentic work](https://postcutoff.com/e/2026-09-02-meta-muse-spark-1-3/index.md)
- 2026-08-05: [Meta launches Muse Code terminal coding agent powered by Muse Spark 1.2](https://postcutoff.com/e/2026-08-05-meta-muse-code-spark-1-2/index.md)
- 2026-04-08: [Meta Superintelligence Labs debuts Muse Spark, its first model](https://postcutoff.com/e/2026-04-08-meta-muse-spark/index.md)
- People: [Mark Zuckerberg](https://postcutoff.com/person/mark-zuckerberg/), [Alexandr Wang](https://postcutoff.com/person/alexandr-wang/)
- Models: [Muse Wonder Full](https://postcutoff.com/m/muse-wonder-full/) (`muse-wonder-full`), [Muse Wonder](https://postcutoff.com/m/muse-wonder/) (`muse-wonder`)
- Archived: 9 posts and 3 videos, listed in https://postcutoff.com/e/2026-10-12-meta-muse-wonder/index.json