H Company has released Holo3.1, a new family of Vision-Language Models (VLMs) designed for computer use agents. These models, based on the Qwen 3.5 family, range in size from 0.8B to 35B parameters and support web, desktop, and mobile environments. Holo3.1 introduces native function-calling support and offers optimized quantized checkpoints for local deployment, enhancing flexibility and cost efficiency. AI
IMPACT Expands options for local deployment of vision-language models for agent tasks.
RANK_REASON Release of a new family of fine-tuned models based on an existing base model, not a novel frontier model release. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
