Meta's Muse Spark rumored to mirror Kimi-K2-style MoE: 816B total at 42B active, 62 layers
scaling01 · x · 2026-10-16
Account scaling01 claims Meta's new Muse Spark model closely matches an architecture he predicted over a month ago: a Kimi-K2-style sparse MoE. A spec sheet shows Muse Spark 1.1/1.2/1.3 at roughly 0.9–1.01T total params with 33.9–36.1B active and 56–58 layers, while the final Muse Spark is 816B@42B across 62 layers — "close enough," he quips, suggesting Meta may have borrowed Kimi-K2's design. Unverified community analysis.
More from Models
- Dev Builds Haunted Escape Room to Test Reflection AI's Beam, Solves It in 30 Seconds — wandb · 2026-10-16
- Claude Opus 5.5 fast mode exits preview: 2.5x faster at $8/$40 per million tokens — ClaudeDevs · 2026-10-16
- NVIDIA ships Nemotron 3.5 Bolt: 74B-A9B model in 24GB of memory — NVIDIAAI · 2026-10-16
- User's Opus 5.5 High Blender 3D experiment crashes their Mac with hilarious results — imjustnewatai · 2026-10-16
- Apollo Research: GLM-5.3 Spontaneously Reasons About How Its Actions Shape Future Training — jd_pressman · 2026-10-16
- Datalab's Chandra 2.5 hits API: 40% faster, more accurate, still $3 per 1,000 pages — VikParuchuri · 2026-10-16