Skip navigation
2/It's a step up on reasoning and coding.
0:00 / 1:14

2/It's a step up on reasoning and coding.

Alexandr Wang

@alexandr_wang764K followers

Visit on X
307 likesDownloadOriginal post

35,480 views

2/It's a step up on reasoning and coding. compared to Muse Spark 1.2, Muse Spark 1.3 wastes fewer turns, uses ~20% fewer tool calls, ~25% fewer tokens, and holds onto requirements well during long-horizon tasks.

6 Comments

Comments from the original post on X (Twitter)

  • Alexandr Wang's profile picture

    @alexandr_wang

    1/ today we’re releasing muse spark 1.3—available in muse code & the meta model api. this is our most capable model yet—frontier performance almost too cheap to meter. much stronger at agentic and coding with better usability. we think users will really notice the jump.

    3,615 likes

  • Alexandr Wang's profile picture

    @alexandr_wang

    3/ available now in Meta Model API and Muse Code. we are cooking! bigger and more capable muse models are close behind 🍉, along with Muse Spark open weights release, and exciting new products. more here →

    271 likes

  • Vector's profile picture

    @PrasVector

    Thats a solid upgrade. Fewer tool calls and tokens usually means less wasted time, so Im curious how it feels in real use. Coding especially needs that efficiency.

    1 like

  • Mustapha's profile picture

    @ElouardiMusta

    I can't wait to test it soon

  • The transfer Account's profile picture

    @TheTransfe69558

    this is observed directly through my usage, amazing work by engineers

  • Hamza 🌙's profile picture

    @ham_zax

    this is the metric I want every coding-model launch to expose: accepted task / tool call accepted patch / token +2 benchmark points is nice. ~20% fewer tool calls + ~25% fewer tokens can matter more if acceptance stays flat. agent loops pay for wandering.