Will there be no next Meta Muse Spark model release by October 31, 2026?

- [2026-04-08] Muse Spark launched on meta.ai and the Meta AI app, with API access in private preview. Meta said then that larger models were in development. https://ai.meta.com/blog/introducing-muse-spark-msl/ - [2026-07-09] Muse Spark 1.1 launched, with the Meta Model API in public preview. https://ai.meta.com/blog/introducing-muse-spark-meta-model-api/ - [2026-08-05] Muse Spark 1.2 and Muse Code (beta) launched, available through the Meta Model API. https://research.meta.ai/blog/introducing-muse-code-and-muse-spark-1-2 - [2026-08-10] Muse Glimmer (30B, Apache 2.0) was released. Meta committed to an open-weight Muse Spark 1.2, which is excluded by the question as a re-release below 1.4. https://futurumgroup.com/insights/meta-reopens-its-models-is-this-a-pc-play-or-a-policy-play/ - [2026-09-02] Muse Spark 1.3 launched in Muse Code and the Meta Model API, with max reasoning. Release cadence so far: 1.1 → 1.2 took about 4 weeks, and 1.2 → 1.3 about 4 weeks. Meta's post says the roadmap includes "bigger models, the Muse Spark open weights release, and more." https://research.meta.ai/blog/introducing-muse-spark-1-3 - [2026-10-05] A third-party guide, updated that day, lists exactly four Muse Spark versions shipped (1.0–1.3). It shows no 1.4. https://codersera.com/blog/muse-spark-complete-guide-2026/ - [2026-10-12] Meta released Muse Wonder and Muse Wonder Full. TechCrunch calls them its first frontier models, described as larger than the "lightweight" Muse Spark. Wonder is on the Meta Model API at $4/$20 per million tokens. Wonder Full is for Muse Code subscribers only. Neither has "Muse Spark" in its name, so neither qualifies. https://techcrunch.com/2026/10/12/meta-releases-its-first-frontier-model-muse-wonder/ - [2026-10-12] Implication: Meta's new top-tier effort is now a separate "Wonder" line. That may reduce the near-term priority of a Spark 1.4, or leave Spark as the cheaper tier. I found no source saying which. - [2026-10-13] The Wikipedia Muse Spark page was last seen listing 1.1 as the headline release and discussing open weights for 1.2. It shows no 1.4. https://en.wikipedia.org/wiki/Muse_Spark - [2026-09-09, approx.] A Manifold market on "Muse Spark 1.4+ release date" exists, in the search snippet marked resolved. I could not verify its state or dates. https://manifold.markets/Bayesian/muse-spark-14-meta-release-date - Same snapshot: the Meta research site's model menu lists Muse Spark 1.1 and 1.2 and Muse Image. It has no 1.4 entry. https://research.meta.ai/blog/introducing-muse-spark-1-3 BOTTOM LINE: As of 2026-10-13 no Muse Spark 1.4+ has been announced or released. The last release, 1.3, came 41 days ago, which is longer than the roughly 4-week gaps before it. Meta's 2026-10-12 launch of the separately named Muse Wonder doesn't count, and it may have taken over the near-term flagship slot. A Spark 1.4 by 2026-10-31 would have to be announced within about 18 days. NOT FOUND: Any official Meta statement or leak of a Spark 1.4 or Spark 2 date. Any scheduled Meta event before 2026-10-31. Meta's Wonder blog post itself (only TechCrunch's account). Any Meta statement on whether Spark continues alongside Wonder. The verified state of the Manifold market.

- [2026-04-08] Muse Spark, the first MSL model, launched. Meta calls it the "first step" of its scaling ladder (TechCrunch). https://techcrunch.com/2026/04/08/meta-debuts-the-muse-spark-model-in-a-ground-up-overhaul-of-its-ai/ - [2026-07-09] Muse Spark 1.1 and the Meta Model API launched, 3 months after the original. Axios reported Meta promising a "significantly more powerful version of Spark later this year" that uses much more compute. https://www.axios.com/2026/07/09/meta-ai-spark-model-update-developer - [2026-08-05] Muse Spark 1.2 and Muse Code launched, about 4 weeks after 1.1 (per orcarouter; the date is also on Meta's research page). https://www.orcarouter.ai/blog/meta-muse-spark-1-2-explained - [2026-09-02] Muse Spark 1.3 launched, about 4 weeks after 1.2. This is the fourth Spark release in five months. Gaps between releases: about 13 weeks, then 4 weeks, then 4 weeks. https://research.meta.ai/blog/introducing-muse-spark-1-3 - [2026-09-02] Meta's own post said "Stay tuned for more updates soon, including bigger models, Muse Spark open weights". Another source says Zuckerberg teased a "🍉"-codenamed release, "coming soon" (secondary source). https://x.com/AIatMeta/status/2095234385129963666 - [2026-10-13] On the 4-week cadence, 1.4 would have been due around Sept 30. As of today it is 41 days since 1.3 with no 1.4. The window to Oct 31 would stretch the gap to 59 days. https://ai.meta.com/research/ - [2026-10-12] Meta's research page lists "Introducing Muse Wonder" (Oct 12), after Muse Spark 1.3 (Sep 2). The page text I could fetch showed no Muse Spark 1.4 item. https://ai.meta.com/research/ - [2026-10-12] A Manifold market on "Muse Spark 1.4+ release date" resolved the "before Sep 21", "before Oct 1" and "before Oct 11" strikes NO. Later strikes show N/A. Comments debate whether the "watermelon" new model family counts. The creator said watermelon counts there, but one trader says it is a new family and not Muse Spark. https://manifold.markets/Bayesian/muse-spark-14-meta-release-date - [2026-10-13] The Polymarket rules here need "Muse Spark" in the name, so a differently named family such as watermelon or Wonder would not qualify. Whether the watermelon model carries the Muse Spark name is unknown. https://manifold.markets/Bayesian/muse-spark-14-meta-release-date - [2026-10-13] OpenCode data lists a "muse-spark-1.4-contributor" model ID with 0 tokens and 2 sessions. The page shows release "Unknown". This may be a placeholder or a leak, but it is not a public launch. The rules exclude labeling errors and placeholders. https://opencode.ai/data/meta/muse-spark-1-4-contributor - [2026-09-27] A user tweet lists "Muse Spark 1.4" as a pending wishlist item. That shows 1.4 was not out as of Sept 27. https://x.com/_simonsmith/status/2104932310692807065 - [2026-09-08] Meta launched a Muse personal agent. Meta also shipped Muse Image (Jul 7), Muse Glimmer (Aug 10) and Muse Wonder (Oct 12). Meta's release rate is high, but many releases are non-Spark and do not qualify. https://imisofts.com/blog/meta-superintelligence-labs/ BOTTOM LINE: Meta shipped Spark point releases at roughly 4-week intervals (Jul 9, Aug 5, Sep 2), so 1.4 looks overdue, and an Oct 13–31 release fits the cadence. But the cadence has slipped, and Meta has said bigger models and a possibly differently named "watermelon" release are coming, which would not qualify unless it is named Muse Spark. I found no official 1.4 announcement as of Oct 13. NOT FOUND: any official Meta statement of a 1.4 date, any credible report of an imminent Spark 1.4, and a full read of the Meta research page. I also could not confirm whether the "watermelon" model will be branded Muse Spark.

YES means: no Meta model named "Muse Spark" and versioned 1.4 or higher (including Muse Spark 2) is made publicly available between now and 2026-10-31, so the market resolves "no next Muse Spark release by Oct 31". - [2026-10-12] The "🍉" model that Zuckerberg teased on Sept 2 ("Next up 🍉") was Muse Wonder. Bloomberg confirms Wonder is "Watermelon" internally. Meta's Sept 2 post named bigger models as the next item, and that slot is now filled by a differently named line. Zuckerberg called Wonder "a much larger and more advanced class" than Spark and described Spark as the "first family." The Manifold dispute over whether "watermelon" counts is moot here, because the Polymarket rules require "Muse Spark" in the name. https://www.business-standard.com/amp/technology/artificial-intelligence/meta-says-new-muse-model-reaches-ai-frontier-after-aggressive-push-126101300103_1.html ; https://ijr.com/discover/tech-2026-10-12-meta-releases-muse-wonder-claims-ai-benchmark-gains-d73b5e62 - [2026-10-12] Meta's own post says Wonder builds on Spark and is "the next step on our scaling ladder." Its stated plan is to use Wonder to "develop specialized models" for Muse users. That implies future efficient derivatives are built from Wonder, and they may carry another name. Meta did not say Spark will continue with 1.4. https://research.meta.ai/blog/muse-wonder - [2026-10-13] Wang said MSL has begun work on "larger and more scaled models." He also said Meta is building cheaper, specialized versions of Wonder for mass deployment. That is where engineering effort is pointed. The efficient tier Spark used to fill is being redefined as Wonder-derived. Nothing in the interview mentions a Spark 1.4. https://www.business-standard.com/amp/technology/artificial-intelligence/meta-says-new-muse-model-reaches-ai-frontier-after-aggressive-push-126101300103_1.html - [2026-10-12] Third-party reporting treats Wonder as the successor to Spark. Fello AI calls it "the successor class to Muse Spark," with Spark last updated at 1.3. Agent-one says it "replaces Muse Spark." Both are secondary sources and not Meta statements, but they show the market reads Spark as superseded. https://felloai.com/muse-wonder/ ; https://www.agent-one.dev/blog/muse-wonder - [2026-10-12] Artificial Analysis shows Wonder at 52 and Wonder Full at 56 on its Intelligence Index, against Spark 1.3 at 48. It notes a "release gap of only 6 weeks." Wonder costs about 2.8x more per task. If Meta's best model is Wonder, a Spark 1.4 sitting between 1.3 and Wonder has less room to matter. Meta also has only just shipped Wonder, and the 1.0→1.3 releases, each about 4 weeks after the last, now look like a pre-Wonder cadence. https://artificialanalysis.ai/articles/muse-wonder ; https://x.com/ArtificialAnlys/status/2109697083519447118 - [2026-10-12] Orcarouter's Oct 12 comparison page calls 1.3 the "current cheap checkpoint" and Wonder the "expensive new flagship," with a routing choice between them. It lists no Spark 1.4. Spark is positioned as the cheap tier, with no sign of a successor. https://www.orcarouter.ai/blog/muse-wonder-vs-muse-spark-1-2 - [2026-10-13] The "muse-spark-1.4-contributor" ID on OpenCode has 0 tokens, 0 users and 2 sessions over Aug 18 to Oct 12. That is too thin for a live model. It is probably a stray or placeholder ID. The market rules exclude placeholders and labeling errors. https://opencode.ai/data/meta/muse-spark-1-4-contributor - [2026-10-13] The only "Muse Spark 1.4" artifact elsewhere is a Solana memecoin created in Sep 2026 on StonkFun. It has a $3.7K market cap and 5 holders, and it is unrelated to Meta. It is a noise item, not a leak. https://phantom.com/tokens/solana/D5UDZ9aftzxfi6D2QnfbENpi3ySf14h3Zf3kPuT1K4Nc - [2026-10-13] The Spark open-weights release is still pending. Fello AI and GCN say the Spark 1.2 weights promised in August have not shipped. Meta's Hugging Face org lists only Glimmer 30B. An open-weight Spark 1.2 is excluded from this market. Pending open-weight work and Wonder's rollout to Muse subscribers may take MSL's release effort ahead of any 1.4. https://felloai.com/muse-wonder/ ; https://gcn.com/meta-opens-muse-glimmer-billion-parameter/21144/ - [2026-10-12] Wonder and Wonder Full are API/Muse Code only so far, and Wonder Full is not on the API. Meta still has consumer rollout to do for Muse ("soon power experiences in Muse"). A Spark 1.4 within 18 days would sit on top of that. https://about.fb.com/news/2026/10/introducing-muse-wonder-models/ - [2026-10-12] The Manifold "Muse Spark 1.4+" market resolved on Oct 12. Its strikes up to Oct 11 resolved NO, and the later strikes resolved N/A. Its creator's call on "watermelon" is unclear in what I could see. A trader there argued Wonder is a new family and "not a muse spark model." https://manifold.markets/Bayesian/muse-spark-14-meta-release-date BOTTOM LINE: The strongest YES (no release) case is that Meta's next big release, the one Zuckerberg teased on Sept 2, was Wonder rather than Spark 1.4. Wang is now talking about larger models and Wonder-derived efficient models, with no Spark 1.4 mentioned in any Meta statement I found. The path to NO still exists, as a surprise 1.4 on the old 4-week-style cadence, but I found no evidence for it. NOT FOUND: Any Meta statement, leak or roadmap line naming Spark 1.4 or Spark 2. Any statement on whether the Spark line continues after Wonder. Any Meta event scheduled before Oct 31. The "specialized models" Meta plans from Wonder have no stated names or dates.

YES means: Meta does NOT make a Muse Spark 1.4+ model publicly available by 2026-10-31. A "Muse Spark 1.4" or higher, or "Muse Spark 2", must be launched publicly. If none is, the market resolves YES. - [2026-10-12] Meta's own Muse Wonder announcement calls Wonder "built on the groundwork we laid with Muse Spark". It names Wonder and Wonder Full as the "next step on our scaling ladder". It sets up Wonder as a separate series, not a Spark version. It says "much more to come" and does not mention a Spark 1.4. https://about.fb.com/news/2026/10/introducing-muse-wonder-models/ - [2026-10-12] Meta's own comparison for Wonder Full uses the "Muse Spark 1.3 trend" as the baseline. That makes 1.3 the reference point for the old line, and Wonder the successor, with no 1.4 in view. https://www.helpnetsecurity.com/2026/10/13/meta-muse-wonder-full-models/ - [2026-10-12] Wang told Bloomberg that Wonder is larger and costlier than the "smaller Muse Spark models". Meta is now building "specialized versions" of Wonder, meant to keep much of its capability while being cheap enough to serve at scale. This is the likeliest route to a cheaper tier. Nothing says that tier will be branded Spark. It may take a Wonder-family name instead, and it was described only as in development. https://www.business-standard.com/amp/technology/artificial-intelligence/meta-says-new-muse-model-reaches-ai-frontier-after-aggressive-push-126101300103_1.html - [2026-10-12] Wonder's blog post says Wonder will be used "in the future" to develop specialized models for everyone using Muse. That implies distillation from Wonder, which is a longer training and evaluation pipeline than a point release. Meta gave no date. https://research.meta.ai/blog/muse-wonder - [2026-10-12] Bloomberg and Axios both say the model Meta shipped under the Wonder name is the "Watermelon" model. So the long-teased next big release has now shipped without the Spark name. The Sept 2 "Spark and bigger models" teaser has therefore been fulfilled by a non-qualifying name. https://www.axios.com/2026/10/12/meta-new-ai-models-anthropic-openai - [2026-10-12] Meta's research page lists only Wonder (Oct 12), Spark 1.3 (Sep 2), Code and Spark 1.2, and Spark 1.1 under "What's new". It shows no 1.4 and no teaser of one. Wonder now sits at the top of the page. https://ai.meta.com/research/ - [2026-10-12] Zuckerberg says Meta is "already working on the next step up", and that future iteration speed will be "much faster". Meta's effort is aimed at larger models, so engineering attention is probably on Wonder-line follow-ups, not Spark point releases (secondary reports). https://www.kucoin.com/news/flash/meta-launches-muse-wonder-ai-model-signals-larger-future-investments - [2026-10-12] Third-party coverage says Spark 1.3 was the last Spark update, in September, and calls Wonder the "successor class" to Spark. It also says the Spark open weights promised in September have not shipped. https://felloai.com/muse-wonder/ - [2026-10-13] A Spark-named release that does ship before Oct 31 might be only the open-weight Spark 1.2 or 1.3, which the rules exclude by name. The pending open-weights item is the likeliest Spark-branded event. It would not qualify. https://felloai.com/muse-wonder/ - [2026-10-13] The only "1.4" artifacts found are not launches. OpenCode lists a "muse-spark-1.4-contributor" ID with 0 tokens, 2 sessions and an unknown release date. A Solana memecoin named Muse Spark 1.4 has a $3.7K market cap and was created in September 2026. Neither is an official model, and the rules exclude placeholders. https://opencode.ai/data/meta/muse-spark-1-4-contributor - [2026-10-12] The Manifold market resolved its "before Oct 11" strike NO. Its later strikes, before Oct 21 and before Nov 1, were marked N/A. A trader argued that Watermelon is a "new model family" and not Spark. That fits Meta's later naming. https://manifold.markets/Bayesian/muse-spark-14-meta-release-date - [2026-10-13] The deadline leaves 18 days. Wonder just shipped, and Meta usually spaces its major launches by weeks. Spark would need its own slot soon after it. There is also an unfinished Wonder rollout to Muse subscribers and a pending Wonder safety report ("soon" per Meta). Those compete for the same launch and comms capacity. https://www.helpnetsecurity.com/2026/10/13/meta-muse-wonder-full-models/ BOTTOM LINE: Meta's flagship attention has moved to the separately named Wonder line. The pending Spark-branded items are the open weights (excluded) and possible cheaper Wonder-derived tiers (name unknown). A qualifying Spark 1.4+ in an 18-day window looks less likely than the old four-week cadence suggested. NOT FOUND: Any official Meta statement or leak of a Spark 1.4 or Spark 2 date. Any Meta statement on whether the Spark line continues alongside Wonder. The name of the planned "specialized" Wonder-derived models. Any scheduled Meta AI event before Oct 31.