SyzygyResearch has released Mach-1-Additive-35B, a 35 billion parameter Mixture of Experts (MoE) model, in GGUF format. This model is designed for efficiency, capable of running on mobile and edge devices with low memory requirements, and achieves up to 120 tokens per second on consumer laptops. The release includes custom GGUF files and a modified llama.cpp fork for optimized performance. AI
IMPACT This release offers a highly efficient 35B MoE model suitable for resource-constrained environments, potentially broadening the accessibility of advanced AI capabilities.
RANK_REASON Release of a specific model with performance metrics and availability on platforms like Hugging Face. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
