AI Briefing
KO

Syzygy Releases Mach-1 35B

·2026.08.20 16:52

Key point

SyzygyResearch has released the 7GB Mach-1-Additive-35B model and a dedicated llama.cpp fork.

Details

SyzygyResearch has released the GGUF version of the Mach-1-Additive-35B model and a custom llama.cpp fork on Hugging Face and GitHub.

This model compresses a 35B MoE architecture into 7GB, optimized for execution on mobile devices, edge devices, and low-memory systems. It has been confirmed to achieve inference speeds of up to 120 t/s in consumer laptop environments.

  • Model: Mach-1-Additive-35B (GGUF)
  • Size: 7GB
  • Inference Speed: Up to 120 t/s (on consumer laptops)
  • Distribution: Hugging Face and GitHub (SyzygyResearch/llama.cpp-mach1)
llminferencehuggingfaceopen-sourcellama-cppedge-ai
Read originalShare on Threads

This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.

Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.

Related

RedditOpen-Weight CLM Released as Jev Alternative, Agent Decision-Making 4x FasterRedditMica v0.1 4B Released: Open-Source Decision Model for Agent Loops Trained for Under $30Hacker NewsQwen3.5-based Small Decision Model 'Kev' Released with TypeSafe API Compatibility and Performance MetricsPyTorchKRPostgreSQL Extension pgjev Released, Supporting Natural Language Filtering Without EmbeddingsPyTorchKRvLLM Releases 'Decision 1.0', AI Models Specialized for Decision-MakingHuggingFace BlogLiquid AI Releases LFM2.5-VL-DSpark: An Experimental Draft Model for Faster Vision-Language Inference