Meta's multimodal reasoning model built for agentic coding and computer use — with a 1M-token context window and parallel sub-agents, now open to developers on the Meta Model API.
Last verified · 2026-07-11 · by Moe Ameen
Muse Spark 1.1 is Meta's multimodal reasoning model, launched on July 9, 2026 by Meta Superintelligence Labs, the group led by chief AI officer Alexandr Wang. It is an upgrade to the original Muse Spark, which shipped in April 2026 as the lab's first proprietary model and Meta's move away from the open Llama line. Version 1.1 adds major gains in tool use, computer use, coding, and multimodal understanding, and it is Meta's entry into the crowded agentic-coding race against Anthropic's Claude and OpenAI's GPT models.
The design center is long-running, agentic work rather than one-shot chat. Muse Spark 1.1 carries a 1 million-token context window with active context management, so it can hold a large codebase or a long session in view; it can act as a main agent or a sub-agent and delegate execution across parallel sub-agents; and it does computer use — operating across multiple applications and maintaining context through extended sessions. Meta trained it to choose the cheapest path at each step: write a script when automation is faster, click when direct interaction is simpler, and batch actions where it can. On the coding side, Meta highlights diagnosing and fixing complex bugs, implementing features in enterprise-grade systems, and running large code migrations.
The multimodal side is where it touches creative work. Muse Spark 1.1 reads images, video, and PDFs, with stated strengths in visual-to-code artifact generation (turn a screenshot or mockup into working UI), ultra-descriptive image and video captioning, and agentic workflow execution. You reach it two ways: through Thinking mode in the Meta AI app and on meta.ai, and through a new Meta Model API in public preview for developers. API pricing is $1.25 per million input tokens and $4.25 per million output tokens — aggressive pricing near lighter tiers like Claude Haiku 4.5, even though Meta positions the model itself as a frontier competitor to GPT-5.5, Claude Opus 4.8, and Gemini 3.1 Pro — with free starter credits for new API accounts.
The honest framing: Muse Spark 1.1 is a reasoning-and-coding model, not a content-production tool. It can draft copy, write a script, caption a video, or reason over a document, but it hands back raw text or code in a chat window or an API response. It does not keep a brand voice across a week of posts, render a carousel or an avatar video, size anything per platform, or schedule and publish to your social channels. Treat the specific benchmarks Meta cited, the preview status, and the exact pricing as a launch-window snapshot — verify current numbers on Meta's own pages.
Muse Spark 1.1 is a strong drafting-and-reasoning brain, and its multimodal captioning is genuinely useful — hand it a rough clip and it will describe the scene in detail or spit out a caption. But everything it returns is a block of text or code in a chat window or an API payload. It writes one caption; it does not write a caption sized for a Reel, a different one for LinkedIn, and a third for X, each in your voice. It describes a video; it does not cut that video into vertical clips, burn in animated subtitles, or reframe it. It drafts a script; it does not turn that script into a face-locked talking-head video and post it. Kompozy is the layer that takes the model's raw output the rest of the way. Paste a Muse Spark-drafted script into Kompozy and it becomes a Persona Short or a HeyGen avatar video with a consistent, brand-owned face; paste a drafted outline and it fans into a Blog Article, an Email Newsletter, a brand-exact Carousel, and native Text Posts — all held to one voice by your Persona Brief and banned-word filters so the volume still reads as your brand, not generic model text.
The division of labor is clean: use Muse Spark 1.1 (or any frontier model) for the thinking, and let Kompozy own the making and the shipping. Kompozy generates the formats a chat model can't — Persona Shorts and Persona Frames, Clipped Shorts from long-form, Quote Graphics, Photo Posts, and Infographics — then schedules and publishes the whole package across nine social platforms plus blog and email from one queue, with Autopilot and a per-post review pipeline. One idea drafted in Muse Spark becomes 25–35 finished, on-brand outputs in Kompozy, live everywhere. The model gives you the words; Kompozy gives you the content operation.
Muse Spark 1.1 is Meta's multimodal reasoning model, launched July 9, 2026 by Meta Superintelligence Labs. It is an upgrade to the original Muse Spark (April 2026), built for agentic coding, tool and computer use, and multimodal understanding, with a 1 million-token context window and parallel sub-agents. It's available in Thinking mode in the Meta AI app and via the new Meta Model API.
On the Meta Model API public preview, Muse Spark 1.1 is priced at $1.25 per million input tokens and $4.25 per million output tokens, with free starter credits for new API accounts. That pricing sits near lighter tiers like Claude Haiku 4.5, even though Meta positions the model itself as a frontier competitor to GPT-5.5, Claude Opus 4.8, and Gemini 3.1 Pro. Confirm current numbers on Meta's pricing page.
No. Muse Spark 1.1 drafts text and code and can caption or summarize media you feed it, but it returns raw output in a chat or API response. It does not render video or carousels, keep a brand voice across a week, size posts per platform, or schedule and publish anywhere. A content engine like Kompozy turns its drafts into finished, on-brand formats and publishes them across platforms.
Meta positions it primarily as an agentic-coding model — diagnosing and fixing complex bugs, implementing features in enterprise systems, and running large code migrations, aided by a 1M-token context window and sub-agent delegation. It enters a market where Anthropic and OpenAI shipped agentic-coding models earlier, and early coverage reads its strength as tool use and orchestration more than outright coding-accuracy leadership — with aggressive pricing as a further draw.
Draft the script, caption, or outline in Muse Spark 1.1, then bring the text into Kompozy. Kompozy renders it into Persona/avatar video, carousels, blogs, newsletters, quote graphics, and text posts — kept on-brand by your Persona Brief — and schedules and publishes the set across nine social platforms plus blog and email.