AI Personal Assistants: Muse, Grok Bot, Cue Usage & Comparisons

A tester found Muse Spark 1.3 produced only a basic endless runner on a simple prompt, improved when asked to research mechanics first, but still missed features Gemini 3.8 Flash added in one shot.

Muse Spark 1.3 scores slightly higher on the Intelligence Index at 61 versus 59, while Gemini 3.8 Flash is faster and cheaper, and Meta says 1.3 uses about 20% fewer tool calls than 1.2.

A Japanese user says Meta Muse may work for inpainting edits; the results are not yet polished but the approach looks viable.

An Arabic post breaks down the 8 free OpenCode models in Hermes, listing Muse Spark 1.3 as the top pick for coding, long agentic tasks and codebase understanding.

Users in China say Meta Muse Spark 1.3 is now free to try on OpenCode and runs very fast, though a VPN is required.

A demo video of a voxel-style "King Kai Planet" generated with Meta Muse Spark 1.3, showing its image and video output.

A user shares a screenshot calling Meta Muse Spark 1.3 a solid model and asks when the Max version will arrive.

A user notes with surprise that Meta's official Muse Spark 1.3 demo first shows the model controlling Notion, and links the developer page.

A 1:33 screen recording shows Meta Muse Spark 1.3 (low) clicking through Wikipedia from the Ferry article to Rihanna.

A user ran the generative octopus test on Muse Spark 1.3 and found the result clearly better than version 1.2.

A user shares a screenshot of switching to the free Muse Spark 1.3 in hermes and compares it with the gemini-3.8-flash-high model they had been using.

Artificial Analysis measured Muse Spark 1.3 (max) at 68 in Muse Code, behind only Claude Opus 5, while the xhigh variant scores 64 at the lowest cost per task, $1.72.