See how ChatGPT, Perplexity and Google AI Overviews describe you today
Free AI Visibility AuditMuse Spark 1.1 is Meta's multimodal reasoning model for agentic work, taking text, images, video, audio and PDF documents as input and returning text. It carries a context window of roughly a million tokens, which is large enough to hold an entire document set or codebase in a single request.
MetaVerified 24/09/2026Released 16/07/202601 / decision
Each figure sits next to the middle of the field, so it can be read as dear or cheap, wide or narrow, rather than floating on its own. Compared against the 154 models we track, not against an absolute standard.
Capability
Not measured
No published benchmark scores yet
Input / 1M tokens
$1.25
field median $0.41
Output / 1M tokens
$4.25
field median $1.55
Context
1,049K
field median 518K
02 / overview
Four questions, answered separately, because somebody arrives at one of them rather than at the top of the page.
The job it was built for, and the job it is not for.
Muse Spark 1.1 is a reasoning model built for agentic tasks, meaning long chains of tool calls and multi-step work rather than single question-and-answer turns. It reads across five input types, text, images, video, audio and files, and always answers in text. It is not an image or video generator, and it is not a small fast model for high-volume classification work.
Where it sits in its line, and when a sibling is the better pick.
Muse Spark 1.1 first appeared in July 2026 as a point release in Meta's Muse Spark line. It is the right pick when a job needs mixed media in, long-running agent behaviour and a context window big enough to avoid chunking. It is the wrong pick for simple, high-frequency calls where a cheaper text-only model does the same job for less.
The API, the apps it powers, and what its limits let you do.
Muse Spark 1.1 is reached through Meta's API, where a single call can carry documents, screen recordings, audio and images alongside the prompt. The very large context and unusually high output ceiling mean you can pass a full corpus in and get a long structured document back without splitting the work into passes. That suits agents that read a whole repository or a stack of PDFs, then write a complete report or a full set of file changes in one go.
What changes because this exists, or why it does not.
The combination that matters here is mixed media in, very long context and a long maximum output in the same model. Work that previously meant transcribing audio with one tool, extracting PDF text with another and stitching the results together can now happen in one call. For agent builders it removes a lot of the plumbing that existed only to work around small context windows.
Superseded by Muse Spark 1.2. See the whole line.
03 / evidence
Split by category, so a strong number never hides a thin evidence base. Verified means we read it on the benchmark's own published results; a provider's claim about its own model is shown and labelled rather than dropped.
No published benchmark scores for this model yet.
We publish a score only where we can link the result to where it was published. Until a benchmark result for this model exists in a source we read, this section stays empty rather than being filled with a provider's marketing figure.
How we decide what counts as evidence04 / ledger
Every published row, grouped by category, each compared with the best published score on the same benchmark. 'Is 64% good' is a question nobody can answer; '26 points behind the leader' is one anybody can.
Nothing in the ledger yet.
Each row here carries a score, the benchmark it came from, what the leading model scored on the same test, and a link to the published result. Rows appear as results are published and read.
How we decide what counts as evidence05 / capability
Where this model is strong, and against how many peers. Ranks are against models with evidence in that category, not against everything we track: ranking against models nobody tested would rank who published, not who is better.
No category scores to shape yet.
A category score is the weighted mean of the benchmarks published for it. With no published rows there is nothing to average, and an empty chart drawn at zero would say something false.
How we decide what counts as evidence06 / cost
List API rates as last read from the provider, with the source on every row, plus every change we have recorded since we started tracking it.
| Charge | Price | Unit | Read on | Source |
|---|---|---|---|---|
| Input | $1.25 | per 1M tokens | 2026-09-24 | Check |
| Output | $4.25 | per 1M tokens | 2026-09-24 | Check |
Muse Spark 1.1 sits in the mid range on input cost and charges a little under four times that for output, so the economics turn on how much text you ask it to write rather than how much you feed it. Given the very large context, it is cheap to read a lot and comparatively expensive to generate a lot, which makes it well suited to summarise-and-extract workloads and less suited to bulk long-form generation.
| Workload | Input / month | Output / month | Cost |
|---|---|---|---|
| A small product team | 20M tokens | 5M tokens | $46.25 |
| A busy support assistant | 200M tokens | 40M tokens | $420.00 |
| A document pipeline | 1000M tokens | 100M tokens | $1,675.00 |
List API rates, no caching and no batch discount, which both providers offer and which change the answer a great deal. Treat these as the ceiling, not the bill.
07 / specs
As listed by the provider's own catalogue and re-read every few hours. Anything absent is absent there too.
| Context window | 1,048,576 tokens |
|---|---|
| Maximum output | 943,718 tokens |
| Modalities | text, image, video, file, audio |
| Released | 16/07/2026 |
| Status | Current |
| Catalogue identifier | meta/muse-spark-1.1 |
08 / lineage
What this model replaced, what replaced it, and what else its provider has in the field.
Also from Meta
09 / line
Every model in this family in release order, so a page from eight months ago says in one glance that two newer ones exist.
Came before
The first we track in this line.
Came after
Muse Spark 1.2 →Ordered by release date and worked out from the naming, so a new member slots in as soon as its page exists. A retired model keeps its page and its place in the line.
10 / notes
What this model changes for a brand trying to be cited in AI answers, and every change we have logged since it launched.
What it changes for you
A model that can read video, audio and PDFs directly means your product demos, webinars and documentation are all now readable source material, not just your web pages. If a SaaS brand keeps its claims locked inside gated PDFs or unlabelled video, agentic models built on Muse Spark 1.1 can still parse them, but only when they are reachable. The practical move is to make sure every format you publish states the product name, the category and the proof plainly, rather than relying on the HTML pages alone.
Most buyers meet Muse Spark 1.1 indirectly, through the API, inside Meta's own assistant surfaces and inside third-party products that have wired it in as their reasoning engine. Because it accepts video and audio, it tends to show up in tools that analyse recorded calls, demos and support sessions, so a buyer may be reading its output without seeing the model name.
Nothing published here yet. Changes appear within hours of a provider announcing them.
11 / questions
The things people ask about this model, answered from what is on this page rather than from anywhere else.
Models change what gets cited. We measure whether AI answers name your brand or your competitors across every assistant, and show you what to change.