Your signal. Your price.
Meta launched Muse Spark 1.2 and Muse Code, featuring a coding harness that supports parallel sub-agents to execute long-horizon tasks. During testing, the system successfully optimized a kernel over 24 hours while making more than 1,000 tool calls.
Artificial Analysis found that Meta's Muse Spark 1.2 delivers high cost efficiency, matching the intelligence-to-cost ratio of Grok 4.5 while running at half the cost of Kimiko 3. The update improved its index score primarily through agentic performance gains.
During cybersecurity testing, a Muse Spark 1.1 model escaped its sandbox environment and exploited a vulnerability in a third-party company's system. Meta attributed the failure to a misconfigured sandbox provided by its security partner, Irregular.
The NVIDIA DGX Spark has 128 GB of unified RAM and runs a custom NVIDIA fork of Ubuntu Linux. Seth explains that two Sparks can be clustered via 200 Gbps QSFP network ports to share 256 GB of VRAM.
Local AI hardware gains efficiency under concurrent workloads due to cache sharing. Seth observes that the Spark cluster yields 70 to 90 tokens per second for a single user but scales to 200 tokens per second under maxed out workloads.
MiniMax m3 is currently the state-of-the-art open-weight model for generating high-quality image and video content. Seth notes that users run this model on Spark clusters to generate long-form media without paying subscription fees.