On 9 July 2026, Meta Superintelligence Labs announced Muse Spark 1.1. The first major Muse update expanded the model from Meta’s assistant into a developer-facing agent platform with parallel tool use.
This release brief was checked against first-party material on 10 August 2026. The date above is the public announcement date, not the date a repository was created or a third-party provider added the model. Where access or weights arrived later, that distinction is recorded below.
Release record
| Field | Verified detail |
|---|---|
| Announcement | 9 July 2026 |
| Availability or weight release | 9 July 2026 in Meta AI Thinking mode and the public-preview Meta Model API |
| Release type | proprietary multimodal reasoning and agent model |
| Access | hosted Meta AI plus an OpenAI-compatible public-preview API; no released weights |
| Architecture | an undisclosed multimodal frontier model trained for planning, parallel subagents, coding, computer use and context compaction |
| Maximum stated context | one million tokens with active compaction and retrieval |
What changed
Muse Spark 1.1 is Meta’s clearest return to the developer model market in this window. It can orchestrate subagents, choose between scripting and interface actions, inspect audio and video, and maintain a million-token working context. Meta’s accompanying 112-page evaluation report also distinguishes unmitigated model capability from the deployed system, documenting where chemical, biological and cyber risks required layered safeguards.
The practical comparison is therefore not simply whether Muse Spark 1.1 has the largest headline score. Teams need to ask whether its architecture, access terms, latency, tool behaviour and evaluation setup match the workload they actually intend to run. A model can lead one harness while losing on cost, refusal behaviour, multilingual quality or repeatability in another.
Benchmarks worth retaining
| Evaluation | Reported result | How to read it |
|---|---|---|
| Design Arena Agentic Web Dev | release comparison reported competitive frontier placement | Human preference depends on the submitted harness and artifact set |
| CyBench pass@1 | 92.9 | Unmitigated capability result from the evaluation report |
| CyberGym pass@1 | 59.0 | Model-level cyber result; deployed safeguards change reachable behaviour |
These are release-time results, not independently reproduced guarantees. Some general-capability charts use internal benchmarks, and the API preview adds developer-controlled tools that can materially change both task success and risk. Scores should remain attached to the disclosed effort setting, agent harness, tool access, timeout, context-management policy and judge model. Moving a number into a procurement sheet without those conditions creates false comparability.
Architecture and access
Muse Spark 1.1 is described as an undisclosed multimodal frontier model trained for planning, parallel subagents, coding, computer use and context compaction with one million tokens with active compaction and retrieval of stated context. Its access position at verification time is hosted Meta AI plus an OpenAI-compatible public-preview API; no released weights. That wording matters: open weights, source-available weights, an API, a product preview and a research demonstration give adopters very different rights and different levels of reproducibility.
Before deployment, record the exact model identifier or checkpoint, inference stack, quantisation, reasoning setting, region, price schedule and supplier terms. If the release uses a custom licence, read the licence itself rather than relying on the word “open” in launch copy. If it is API-only, preserve the dated documentation and change-notice route because the served snapshot can change without a downloadable artefact.
What an evaluation should test next
For Muse Spark 1.1, a credible internal gate should include:
- a frozen set of representative tasks with pass, fail and abstain criteria;
- a matched baseline using the same tools, timeout, prompt budget and reviewer rubric;
- repeated runs to expose variance rather than reporting a single best attempt;
- latency, token use and total task cost alongside task success;
- adversarial, multilingual and long-context cases relevant to the real deployment; and
- rollback evidence showing the previous model can be restored safely.
The wider model change-control guide explains how to keep model, prompt, tool and corpus changes reconstructable. The AI dependency inventory guide covers the release and supplier records needed after deployment.
AIEngine verdict
Muse Spark 1.1 is a major frontier-agent release. Adopters should reproduce Meta’s distinction between model and system, isolate tools, and log compaction, subagent delegation and policy interventions as first-class events.
This is a launch assessment, not a certification. Benchmark leadership is useful evidence of where to test; it is not authorization to place the model in a high-impact workflow without domain evaluation, security review and an accountable owner.
Primary sources
Image provenance
Hero image: Meta official Muse Spark artwork. The locally served WebP is a crop of the first-party release or model-card asset recorded in the repository provenance manifest.
