Muse Spark 1.4 Contributor Rumor: A Listing, Not Confirmed Access

Evidence record: identity, access and unknowns

Where the rumor comes from

LLMLearner’s September 25 record attributes a September 24 screenshot to X account @kimmonismus. The linked X original could not be retrieved during our review, so that attribution remains secondary. We independently opened the OpenCode page carrying the identifier; that verifies the listing’s existence, not the screenshot’s full history. Secondary attribution OpenCode data page

The Reddit thread asks whether the model is already here and links the same page. Its author, Final_Initial, added that an attempt to use the model did not work. That qualification belongs beside the exciting headline. We could verify the public edit, but not the attempted request, account configuration or reason for failure. It is evidence of a reported problem, not a conclusive test of every possible access route. Public discussion

These references should not be counted as several independent confirmations of a release. They converge on a single third-party listing. Repetition can make a name familiar without adding evidence about the service behind it. We therefore classify the story as an unverified model-identity rumor with an accessible public basis.

What the data page actually tells us

The inspected OpenCode page shows unknown context, output, knowledge, release and input fields. Its activity display has zero tokens and zero unique users, while also displaying two completed sessions. We cannot resolve that mismatch from the page alone, and do not treat those sessions as proof of successful inference. OpenCode data page

A data interface can be useful for noticing a possible model name without being a release authority. In this case, the unresolved fields leave the core product questions open. A reader cannot infer an input limit from a naming convention, or infer support for images from a neighboring model. The sensible interpretation is narrow: this particular interface has a record for this particular string.

Nor does the lack of displayed token usage establish that no one has tested a model privately or elsewhere. It describes the observed window in this interface. Conversely, an empty cost display cannot be read as a free pricing tier. Neither direction supplies a reliable purchasing or deployment decision; both require documentation that connects the identifier to an actual service. OpenCode data page

Why the Contributor suffix matters

Meta’s inspected page identifies Muse Spark 1.3 and distinguishes its contributor variant from the standard variant. The contributor row says it is used to improve Meta’s products; the standard row says it is not. Those are statements about the listed 1.3 variants, not verified terms for the rumored 1.4 identifier. Meta model documentation

Our editorial takeaway is that version and service terms need separate review. If a future endpoint uses the Contributor label, teams should read that endpoint’s documented data-use conditions before submitting proprietary code. A similar spelling does not by itself establish identical retention, training use, account eligibility or contractual protections. This article makes no claim about what a future policy will say.

The distinction also affects comparison design. A model choice is partly a capability decision and partly a decision about acceptable access conditions. A faster or less expensive option would still need to fit the project’s requirements. Evaluate those requirements using the actual service documentation, and keep any version-specific price or quota out of a comparison until its billing unit and scope are clear.

September 30 update: do not equate Watermelon with this listing

A separate Watermelon rumor should not be used to fill the gaps in this listing. Alistair Wren’s OrcaRouter article, published September 6 and subsequently updated, describes Watermelon as a working codename with unresolved product lineage. It attributes an October target to The Information’s August 25 reporting. We verified this secondary account, not the original report or its underlying evidence; October remains a reported target, not a confirmed date. Attributed Watermelon report.

OrcaRouter also relays a September 2 Zuckerberg sign-off through X account @scaling01, mentioning Watermelon and Muse Spark open-weight releases separately. We did not independently retrieve that original. Even as reported, the wording does not establish that Watermelon itself will have downloadable weights, or that it is the muse-spark-1.4-contributor identifier. A future downloadable language model could matter for local deployment, but no license, hardware requirement or matching checkpoint is established here. Keep the two identities separate until a provider document explicitly connects them. This update clarifies the rumor’s implications; it does not upgrade either claim to a confirmed release. Reported wording and limitations.

What would make a 1.4 update important

If this label eventually corresponds to a documented successor, the useful question for coding teams would be whether it improves the work they already assign to a language model. That is a conditional assessment of potential impact, not a prediction of capability or a probability estimate for this rumor.

For a bug-fixing assistant, a useful improvement might mean resolving a reproducible issue without introducing another regression. For repository exploration, it might mean finding the relevant implementation and explaining dependencies accurately. For a tool-using agent, it might mean completing the requested change while respecting the allowed scope. These are proposed evaluation goals; none has been demonstrated for 1.4 in this report.

A version number cannot tell us which of those outcomes will improve, or whether a change would justify migration costs. The rumor matters because it could add another candidate to an evaluation shortlist. It does not yet supply a reason to replace a working endpoint, promise a throughput gain or announce a new quality leader. Potential impact is substantial; evidence of those outcomes remains absent.

How to prepare a useful comparison without inventing results

Prepare a small set of representative tasks and a record of the current model’s settings. Keep the future Muse Spark entry empty until its exact identifier and usable access are established. This is an evaluation plan, not a benchmark that we have run or a claim that the rumored model is competitive. Unresolved listing context

Choose cases with inspectable outcomes: a failing test that should pass, a behavior that must remain unchanged, and a codebase question whose answer can be checked against the repository. Record the input, permitted tools, retry policy and completion criteria. If an eventual service exposes different reasoning controls, document those differences rather than implying that similarly named settings are equivalent.

Cost and speed comparisons should follow the same discipline. Count the work required to obtain an acceptable result, including retries, and keep observations separate from advertised token rates. If the endpoint cannot complete a request, record an access failure rather than assigning a quality score. This avoids turning account restrictions, configuration mistakes or an unconfirmed route into a misleading judgment about model intelligence. Unresolved listing context

The next evidence that would change this report

A provider announcement naming the exact model, endpoint documentation specifying its inputs and limits, or a reproducible public access report would each add different information. None should silently substitute for the others. A roadmap could establish intent before access exists; a working route could still leave pricing or deployment scope unresolved. Unresolved listing context

We will update this page when new evidence changes the status, preserving the original publication date and stating what was corrected. If the listing disappears, that would change the public evidence record without proving a cancellation. If Meta names a different successor, we would separate that confirmed identity from this earlier label instead of retroactively declaring the rumor accurate.

For now, use the source links to distinguish observed page content from commentary, and keep this item on a watchlist rather than a production configuration. Our review was targeted and does not establish an exhaustive account of Meta’s announcements or private plans. No release date, context limit, price, license, supported modality or performance score for Muse Spark 1.4 is asserted here. Unresolved listing context

FAQ