A recent week saw the release of three open-weight AI models, each with distinct characteristics and licensing. MiniCPM-V 4.7, a 35B parameter model from OpenBMB, is notable for its ability to run on mobile devices like the iPhone 17 Pro Max with minimal memory requirements and an Apache 2.0 license, allowing commercial use. Ling-3.1-flash by inclusionAI, a 560B parameter model, operates with 25B active parameters and offers a 1M token context window under an MIT license, with hardware requirements varying based on GPU memory. Muse Spark Open 816B-A42B from Meta Superintelligence Labs is an 816B parameter model supporting multimodal inputs, but its Muse Model License imposes conditions on commercial use and attribution, particularly for model-as-a-service offerings exceeding $20 million in revenue. AI
IMPACT Highlights the critical need to evaluate memory, licensing, and active parameters beyond just total parameter count when selecting AI models for deployment.
RANK_REASON The item details the release and technical specifications of three new open-weight AI models, focusing on their parameters, memory requirements, and licensing terms. [lever_c_demoted from research: ic=1 ai=1.0]
- Apache 2.0
- GB300
- H200
- inclusionAI
- iPhone 17 Pro Max
- Ling-3.1-flash
- Meta Superintelligence Labs
- MiniCPM-V 4.7
- MIT
- Muse Model License
- Nokka
- Ollama Cloud
- OpenBMB
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
