Meta Superintelligence Labs debuts Muse Spark, its first model
Confirmed
Importance: major (4 of 5)The takeaway
On 2026-04-08 Meta Superintelligence Labs (led by Alexandr Wang) released Muse Spark (code-named Avocado), the first model of the new Muse series and the result of a nine-month ground-up rebuild of Meta’s AI stack.
Status
- Claim
Confirmed
- Our reporting
- High confidence
- Importance
- Major (4 of 5)
- Last verified
- 29 October 2026
Your AI and this story
- GPT-6 AstraIn its training data
- Claude Opus 5.5In its training data
- Gemini 4 Hydrogen8 days after its cutoff
- Grok 4.7In its training data
It is after the cutoff of Gemini 4 Hydrogen, and can be in the training data of GPT-6 Astra, Claude Opus 5.5 and Grok 4.7.
Key facts
- Announced 2026-04-08; first model from Meta Superintelligence Labs; code-named Avocado
- Described as small and fast by design, reasoning in science, math and health; supports parallel subagents
- Powers the Meta AI app and meta.ai at launch; rolling out to WhatsApp, Instagram, Facebook, Messenger and AI glasses
- Private-preview API access for select partners
- Not open weights; Meta said it hopes to open-source future versions
- No numeric benchmarks published in the official post
- Launched with Meta’s Advanced AI Scaling Framework v2 (Apr 8; replaces the 2025 Frontier AI Framework, adds a loss-of-control section, pre- and post-safeguard testing, and Safety & Preparedness Reports)
- Muse Spark Safety & Preparedness Report (159 pages; arXiv May 14, 2026): chem/bio capabilities likely reached the framework’s ‘high risk’ category before safeguards; deployed at ‘acceptable’ residual risk after mitigations; Apollo Research found the highest evaluation awareness it had observed
Show 2 more
- Followed by Muse Image and Muse Video, Muse Spark 1.1 (July), 1.2 (Aug) and open-weight Muse Glimmer (Aug)
- Irregular (Apr 15, first Meta collaboration): Muse Spark solved four of six hard or expert Atomic Tasks at least once but completed no full CyScenarioBench scenario; Irregular agreed with Meta’s Advanced AI Scaling Framework finding that it lacks autonomous offensive capability and ‘does not materially alter the cyber threat landscape’
What happened
Meta released Muse Spark, the first model built by Meta Superintelligence Labs (MSL), the unit formed in 2025 after Meta’s roughly $14B deal with Scale AI that brought in Alexandr Wang. Meta says MSL rebuilt its AI stack from the ground up in nine months. Muse Spark powers Meta AI with reasoning, visual understanding, health Q&A developed with physician input, visual coding (websites, mini-games) and parallel subagents.
Why it matters
It marked Meta’s break from the Llama brand and from default open-weights releases for its frontier model, and was the first test of whether Meta’s enormous 2025-26 talent and capex spending could produce a competitive model.
Sources
7 sources from 6 sites. Numbers match the chips in the text.
7 sources: 5 primary, 2 press
Primary
- Meta - Introducing Muse Sparkabout.fb.com, official
- Meta AI: Scaling how we build and test our most advanced AI (Advanced AI Scaling Framework)ai.meta.com, official
- Muse Spark Safety & Preparedness Report (Meta)ai.meta.com, official
- arXiv 2606.12429: Muse Spark Safety & Preparedness Reportarxiv.org, paper
- Irregular: Assessing Meta’s Muse Spark against offensive security benchmarks (Apr 15)irregular.com, official
Press
Changes
- Added Irregular’s pre-release cyber assessment (read directly Oct 29) (news agent)
- Added Advanced AI Scaling Framework v2 and the Safety & Preparedness Report
- Filed