{"slug": "meta-muse-wonder-for-vibe-coding-and-where-to-host-the-app", "title": "Meta Muse Wonder for vibe coding (and where to host the app)", "summary": "Meta announced Muse Wonder and Muse Wonder Full, two models for personal and professional agents, with Meta's reported results showing Muse Wonder Full leading DeepSWE 1.1 at 79.0, SWE-Marathon at 57.5 and AutomationBench-AA at 45.4, while Claude Opus 5.5 still leads Terminal-Bench 4.0-AA at 59.6, ProgramBench at 70.5 and the overall Artificial Analysis Index at 58. Muse Wonder is available today through the Meta Model API and Muse Code, and Muse Wonder Full is available today for Muse Code subscribers, though Meta published no pricing, general consumer rollout date or Vibe Code Bench score. The figures are Meta's reported results rather than an independent league table.", "body_md": "[Blog](https://oncroft.net/blog/) · October 13, 2026 · Rich Chetwynd\n\n# Meta Muse Wonder for vibe coding (and where to host the app)\n\nMeta announced Muse Wonder with strong reported scores on long-horizon coding and agent workflows. Claude Opus 5.5 still leads some coding benches. After you build with any model, Croft is where your team runs the app.\n\n## Quick answer\n\n**Meta announced Muse Wonder** today: two models, Muse Wonder and Muse Wonder Full, aimed at personal and professional agents. On **Meta’s reported table** (max settings), Muse Wonder Full leads several long-horizon coding and agent-workflow benches — DeepSWE 1.1 at **79.0**, SWE-Marathon at **57.5** (xhigh), AutomationBench-AA at **45.4**. **Claude Opus 5.5** still leads Terminal-Bench 4.0-AA (**59.6**), ProgramBench (** 70.5**), and the overall Artificial Analysis Index (** 58**).\n\nThose figures are **Meta’s reported results**, not an independent league table. For teams and developers [vibe coding](https://oncroft.net/vibe-coding/) an app, the practical split is: pick the model that fits the job; then put the app somewhere private the team can actually run.\n\n## What Meta announced\n\nFrom Meta’s [Muse Wonder research post](https://research.meta.ai/blog/muse-wonder):\n\n- Two models: **Muse Wonder** and**Muse Wonder Full**\n- **Active Memory** (carry relevant context across conversations) and**Active Focus** (keep working a hard goal, try other approaches, check in before important actions)\n- Domain work Meta highlights: law, finance, and health, plus frontier coding\n- **Muse Wonder** is available today through the**Meta Model API** and**Muse Code**\n- **Muse Wonder Full** is available today for**Muse Code subscribers**\n- Meta says Muse Wonder **will soon power experiences in Muse** , rolling out first to subscribers, and will power**artifact generation**\n\nMeta did not publish pricing, a general consumer rollout date, or a Vibe Code Bench score in that post.\n\n## What Meta’s numbers mean for vibe coding\n\nVibe coding is describing an app and steering until it works. Long-horizon coding benches are a useful proxy for that: the model has to keep requirements in mind, change a codebase, test, and recover. Agent and workflow benches matter if the app has to do a job, not just render a screen.\n\nOn **Meta’s comparison table** (Muse Wonder Full max, Muse Wonder max, Claude Fable 5.1 max, GPT-6 Astra max, Claude Opus 5.5 max):\n\n**Where Meta reports Muse Wonder Full ahead**\n\n- **DeepSWE 1.1** (long-horizon coding):**79.0** vs Opus 5.5 at 74.2, Astra at 74.1, Fable 5.1 at 67.4. Muse Wonder (the smaller of the two) is**78.8** .\n- **SWE-Marathon** (ultra-long-horizon engineering):**57.5** (xhigh) vs Opus 5.5 at 46.3. Muse Wonder is**53.1** .\n- **AutomationBench-AA** (end-to-end business workflows):**45.4** vs Astra 41.6 and Opus 5.5 at 41.2.\n- **Harvey LAB-AA** (legal agent work):**13.1** vs Opus 5.5 at 4.2.\n- **MemoryBench** (remembering context): Muse Wonder**85.6** , Muse Wonder Full**85.5** , vs Opus 5.5 at 81.5.\n\n**Where Meta’s table still has Claude Opus 5.5 ahead**\n\n- **Terminal-Bench 4.0-AA** (agentic terminal coding):**59.6** vs Muse Wonder Full at 57.1. GPT-6 Astra is 59.1.\n- **ProgramBench** (from-scratch programming):**70.5** vs Muse Wonder Full at 59.0 — a wide gap on this bench.\n- **Artificial Analysis Index** (overall):**58** vs Muse Wonder Full at**56** .\n- **HealthBench Pro** :**65.6** (Astra 65.3) vs Muse Wonder Full at 62.3.\n- **OSWorld V2.1** (partial):**81.8** vs Muse Wonder Full at 78.4.\n\nA compact slice of the coding and overall scores Meta reported:\n\n| Benchmark | Muse Wonder Full | Claude Opus 5.5 | Lead on Meta’s table | \n|---|---|---|---|\n| DeepSWE 1.1 | 79.0 | 74.2 | Muse Wonder Full | \n| SWE-Marathon | 57.5 | 46.3 | Muse Wonder Full | \n| AutomationBench-AA | 45.4 | 41.2 | Muse Wonder Full | \n| Terminal-Bench 4.0-AA | 57.1 | 59.6 | Claude Opus 5.5 | \n| ProgramBench | 59.0 | 70.5 | Claude Opus 5.5 | \n| Artificial Analysis Index | 56 | 58 | Claude Opus 5.5 | \n\nHonest read for a team building apps with AI: **Muse Wonder looks strong when the work runs long** — multi-step coding, business workflows, keeping context. **Opus 5.5 still wins some of the coding suites that look more like from-scratch programs and terminal agents**, and it leads Meta’s overall index. GPT-6 Astra and Claude Fable 5.1 sit in the mix on the same table.\n\nTreat this as **Meta’s snapshot on launch day**. Public app-building leaderboards (see [best models for vibe coding](https://oncroft.net/best-models-for-vibe-coding/)) measure something different, and they move.\n\n## After you build, you still need a host\n\nWhichever model writes the app — Muse Wonder, Claude, GPT, Cursor, Grok — a chat preview is not a place your team can work. You still need:\n\n- A **private server** that stays up\n- **Company login** , so only your people get in\n- A **database** everyone shares\n- **Schedules** for jobs that should run without someone clicking\n- **Monitoring** , so a bad deploy is visible\n- **Backups and rollback** when a prompt breaks something\n\nThat is hosting. It is also where most vibe-coded prototypes die.\n\n## Croft is the place the team runs it\n\nCroft does not sell models and does not pick a vendor. **Croft is an MCP server**, so it works with **any MCP-capable assistant**. Meta’s post does not describe a Muse-specific Croft integration; use the client you already have.\n\nThe loop is the same:\n\n1. Describe the app in your assistant (or in [Ask Croft](https://oncroft.net/build/) in the browser).\n2. Deploy to your croft.\n3. Invite the team. They sign in once.\n\nYou get a live URL on **your own private server**, behind **one company login**, with a **database**, **schedules**, and **monitoring** included. Connect setup: [Claude](https://oncroft.net/connect/claude/), [ChatGPT](https://oncroft.net/connect/chatgpt/), [Cursor](https://oncroft.net/connect/cursor/), [Grok](https://oncroft.net/connect/grok/), or [any MCP client](https://oncroft.net/connect/).\n\n## Related\n\n- [Best models for vibe coding](https://oncroft.net/best-models-for-vibe-coding/) — public benches, assistants, and how to choose\n- [Vibe coding hosting](https://oncroft.net/vibe-coding/) — where the app lives after the chat window\n- [AI app hosting](https://oncroft.net/ai-app-hosting/) — private hosting for apps you build with AI\n- [Connect your AI assistant](https://oncroft.net/connect/) — MCP setup\n- [Google Gemini 4 Argon for vibe coding](https://oncroft.net/blog/google-gemini-4-argon-vibe-coding/) — the last frontier-model drop, same hosting gap\n\n## FAQ\n\n### What is Muse Wonder?\n\nMeta’s new frontier series, launched 13 October 2026 (NZ): **Muse Wonder** and **Muse Wonder Full**. Meta says they target personal and professional agent work, with Active Memory and Active Focus. Availability in Meta’s post: Meta Model API and Muse Code today; Full for Muse Code subscribers; Muse product rollout coming, subscribers first.\n\n### Is Muse Wonder the best model for vibe coding?\n\nOn Meta’s table it is the one to watch for **long-horizon coding** and **agent/business workflows**. Opus 5.5 still leads **Terminal-Bench**, **ProgramBench**, and the **Artificial Analysis Index**. Use the model that matches the job; rankings will move.\n\n### Can I connect Muse Wonder to Croft?\n\nCroft works with **any MCP-capable assistant**. Meta did not publish a Croft (or MCP) setup in the announcement. Point your MCP client at Croft, build, deploy.\n\n### Where should a team host an app built with Muse Wonder?\n\nOn a private host with company login and a database. That is Croft. Public preview URLs are for demos, not internal tools.\n\n### How much does Croft cost?\n\nSolo from $24/month on annual billing, Team $99/month on annual. 7-day free trial, no card to start. [Pricing →](https://oncroft.net/pricing/)\n\n## Where to learn more\n\n- **Meta research post:**[Muse Wonder](https://research.meta.ai/blog/muse-wonder)\n- **Croft:**[Get your croft](https://app.oncroft.net/signup) — 7-day free trial, Solo from $24/mo annual\n\n## Stake out your croft.\n\nYour team's first app could be live before lunch.\n\n[Start free trial](https://app.oncroft.net/signup)\n\n7 days free, no card to start. From $24/month - cancel anytime and take everything with you.", "url": "https://wpnews.pro/news/meta-muse-wonder-for-vibe-coding-and-where-to-host-the-app", "canonical_source": "https://oncroft.net/blog/muse-wonder-vibe-coding/", "published_at": "2026-10-12 11:00:00+00:00", "updated_at": "2026-10-12 21:20:29.803020+00:00", "lang": "en", "topics": ["artificial-intelligence", "large-language-models", "ai-agents", "ai-products", "ai-tools"], "entities": ["Meta", "Muse Wonder", "Muse Wonder Full", "Claude Opus 5.5", "Meta Model API", "Muse Code", "GPT-6 Astra", "Claude Fable 5.1"], "also_reported_by": [], "alternates": {"html": "https://wpnews.pro/news/meta-muse-wonder-for-vibe-coding-and-where-to-host-the-app", "markdown": "https://wpnews.pro/news/meta-muse-wonder-for-vibe-coding-and-where-to-host-the-app.md", "text": "https://wpnews.pro/news/meta-muse-wonder-for-vibe-coding-and-where-to-host-the-app.txt", "jsonld": "https://wpnews.pro/news/meta-muse-wonder-for-vibe-coding-and-where-to-host-the-app.jsonld"}}