News
Guide
Is AI Making Your Organization More Productive or More Dependent?
Key Takeaways
- Muse Spark 1.1 adds multimodal, coding and agentic improvements.
- Developers can now access the upgraded model via Meta Model API.
- B2B software teams gain advanced automation and context handling for complex workflows.
Meta Superintelligence Labs released Muse Spark 1.1 on Thursday, a multimodal reasoning model built for agentic AI tasks. The upgrade targets gains in tool and computer use, coding and multimodal understanding, according to Meta.
Alongside the model, Meta opened a public preview of the Meta Model API, giving developers direct access to Muse Spark 1.1. The model is also available in "Thinking" mode in the Meta AI app and at meta.ai.
"Meta is clearly building for serious agentic coding — strong tool use at a price point that makes it viable to run real coding workloads at scale. That combination is rare, and it's exactly why we wanted Cline developers to have access early."
- Saoud Rizwan
CEO, Cline
Muse Spark 1.1 Feature Breakdown
Key capabilities Meta claims for Muse Spark 1.1 include:
| Key Feature | How It Works |
|---|---|
| Agentic orchestration | Orchestrates multi-agent systems with parallel subagent delegation |
| 1M-token context window | Manages context, retrieves earlier work and compacts steps |
| Computer use | Automates desktop workflows across multiple apps via scripts or clicks |
| Coding improvements | Diagnoses bugs, implements features and executes code migrations |
| Multimodal reasoning | Processes images, video and PDFs for grounded, structured outputs |
Agentic AI: What Leaders Need to Get Right
Agentic AI has gone operational, but scaling it demands disciplined architecture, governance and integration.
Beyond Single Models: Multi-Agent Workflows
AI agents have evolved from task-specific assistants into autonomous systems that adapt in real time, solve multi-step problems and complete end-to-end work with limited human oversight.
Open multi-agent frameworks enable dynamic workflows across documents, APIs and enterprise systems — a baseline for deploying agentic AI at scale.
Multimodal capabilities extend the value further. Agentic platforms now transform unstructured content — PDFs, contracts, images and video — into structured, automation-ready data, enabling agents to reason across heterogeneous inputs.
Governance Is an Engineering Problem
As multi-agent systems embed deeper into enterprise infrastructure, AI governance becomes a core engineering concern. Effective agentic AI development draws on cross-disciplinary skills spanning ML and LLM operations, cognitive science, systems architecture, cybersecurity and human-computer interaction design.
These disciplines matter because the opacity problem is growing. A Bank of America software developer warned VKTR that AI-generated codebases and enterprise workflows now create technical debt that even experienced engineers struggle to navigate.
Meta in the News
Meta moved toward major expansion over the past two years, anchored by a $14.8 billion investment for a 49% non-voting stake in Scale AI that brought CEO Alexandr Wang in to lead the newly formed Meta Superintelligence Labs.
The company acquired RISC-V chip startup Rivos to accelerate its MTIA semiconductor roadmap and pursued a $2 billion purchase of Singapore-based AI agent startup Manus — a deal China's Ministry of Commerce blocked, marking Beijing's first formal veto of a US tech acquisition involving a startup with Chinese origins.
Meta also joined the US General Services Administration's OneGov initiative, giving federal agencies access to its Llama models.