LLM Digest
Subscribe

AI Storyline

3 items · 2 sources · 3 days

View as JSON

Operational story trace

Technical Report

Latest change

Clover-35B-A3B-Preview's report (Nov 18) describes a mixture-of-experts diffusion language model with 35B total and 3B active parameters that supports a thinking mode.

Earlier contextThe story so far

This is a loose thread of unrelated model technical reports, not one developing event. It opened with Tencent Hunyuan's Hy-Nomas world-model report and benchmark on Oct 30, then MNN-Pocket-0.3B, a 300M-parameter on-device model with 32K context, on Nov 9.

editor-curated · source-linked

Arc

Oct 30Nov 18 · now
WORLD MODELS · Oct 30
Tencent Hunyuan publishes the Hy-Nomas report and Hy-Nomas-Bench
1 source · show source ▾
ON-DEVICE · Nov 9
MNN-Pocket-0.3B targets on-device deployment at 300M parameters
1 source · show source ▾
NOW · Nov 18
Clover-35B-A3B-Preview: a sparse MoE diffusion LM with thinking mode
1 source · show source ▾

What to watch — open questions

  • Are the Hy-Nomas, MNN-Pocket-0.3B, or Clover weights released under a license that permits commercial use?
  • Does Hy-Nomas-Bench gain adoption outside Tencent's own evaluations?
  • Do mainstream inference servers support Clover's diffusion decoding, or does it need a custom runtime?
  • Do independent evals reproduce MNN-Pocket-0.3B's reported quality at INT4?
How this thread was built
editor wrote the arc · 3 beats

Storylines are threaded mechanically from the feed: stories that share a distinctive anchor across multiple days and sources. Each item links to its original source. The evidence trace, current state, and open questions are written by the editor routine and refreshed whenever a new beat lands.