Meta's Muse Spark 1.3 Is Its Most Powerful AI Model Yet
Meta released Muse Spark 1.3 on September 2, 2026 — its fourth Muse Spark model in five months, and the most capable one yet. It's built for agentic coding: sustaining long, multi-step tasks rather than just answering single questions. Here's what actually changed.
⚡ Quick facts
- Released: September 2, 2026 — Meta's 4th Muse Spark release in 5 months
- Benchmark: Scores 61 on the Artificial Analysis Intelligence Index, level with GPT-5.6 Sol and Grok 4.6
- Efficiency: ~20% fewer tool calls, ~25% fewer tokens than Muse Spark 1.2
- Pricing: $1.25/M input tokens, $4.25/M output tokens — unchanged from before
What Muse Spark 1.3 actually does
Muse Spark 1.3 is designed around agentic coding — handling open-ended objectives that require many steps, not just single prompts. Meta says the model can sustain longer-horizon work by collaborating with a user and juggling multiple workflows in a single, long thread. When given an open-ended goal, it uses tools to gather its own context from messy or conflicting sources, proactively corrects gaps in its own plan, and keeps track of what it has learned along the way to produce a final result.
How it compares to rivals
On the Artificial Analysis Intelligence Index, Muse Spark 1.3 scores 61 — roughly level with OpenAI's GPT-5.6 Sol and xAI's Grok 4.6. What stands out is efficiency: Meta measured about 20% fewer tool calls and 25% fewer tokens used to complete the same tasks compared to the previous Muse Spark 1.2, while keeping it the cheapest strong model per task among that comparison group.
Where you can (and can't) use it
Muse Spark 1.3 ships inside Meta's own Muse Code product and the Meta Model API, at $1.25 per million input tokens and $4.25 per million output tokens, with a cheaper $0.10/$0.20 contributor tier. Unlike some of Meta's earlier open releases, this one's weights are closed — you can't download and self-host it. Its most advanced "max reasoning" mode is also still gated behind further internal safety testing.
Why this matters
This is part of the same rapid-fire pattern this site has been tracking all week — four frontier AI model launches (Claude Fable 5.1, Gemini 3.8 Flash, Muse Spark 1.3, and GPT-6 Astra) landed within roughly 72 hours of each other in early September 2026. For everyday users the direct impact is limited since Muse Spark targets developers, but it's another sign of how fast the major AI labs are shipping right now.
Frequently asked questions
What is Muse Spark 1.3?
Meta's newest, most powerful AI model, released September 2, 2026, focused on agentic coding and long multi-step tasks.
How good is it compared to other AI models?
It scores 61 on the Artificial Analysis Intelligence Index, roughly level with GPT-5.6 Sol and Grok 4.6, while cheaper per task.
What makes it better than the previous version?
About 20% fewer tool calls and 25% fewer tokens used than Muse Spark 1.2, at the same price.
Can I download and self-host Muse Spark 1.3?
No -- its weights are closed. It's only available through Meta's Muse Code and the Meta Model API.
How much does it cost to use?
$1.25 per million input tokens and $4.25 per million output tokens, unchanged from previous pricing.