Latest open artifacts (#22): Zyphra, Cohere, and Poolside are expanding the breadth of the ecosystem

2026-06-28 · Source: Interconnects AI · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Intermediate, short

Summary

The open model ecosystem is experiencing increasing diversity, moving beyond a few dominant players to include a wide range of organizations with varied motivations. These include "Pure" model makers like Zyphra, Cohere, and Poolside, Big Tech firms such as Alibaba and NVIDIA, and product companies like JetBrains. Recent notable releases include NVIDIA's Nemotron-3-Ultra-550B-A55B-BF16, which uses LatentMoE for speed and adopts the OpenMDW license for model weights. Cohere released its flagship Command A+ (218B-A25B MoE) under Apache 2.0, offering multi-modal, multi-lingual, and agentic capabilities. Zai-org's GLM-5.2 continues to impress with its everyday usability, while Zyphra introduced ZAYA1-74B-preview (74B-A4B MoE) trained on AMD GPUs. Poolside also released Laguna-M.1 under Apache 2.0, committing to future open releases. This expanding ecosystem underscores the strength of diverse actors in AI development.

Key takeaway

For Machine Learning Engineers evaluating open-source models, the ecosystem's increasing diversity offers more specialized options beyond frontier models. You should explore niche models like Zyphra's ZAYA1-74B-preview for specific use cases and consider models like Cohere's Command A+ for multi-modal, agentic capabilities, noting its 4-bit quantization for single B200 deployment. Pay attention to emerging licenses like OpenMDW, adopted by NVIDIA, which are tailored for model weights, ensuring proper compliance and usage rights.

Key insights

The open model ecosystem is diversifying with varied actors and motivations, fostering innovation and specialized model development.

Principles

• Open model development is driven by diverse actors and motivations.
• Model-specific licenses like OpenMDW are emerging for weights.
• Specialized, smaller models can meet product needs effectively.

In practice

• Use 4-bit quantization for large MoE models on single B200 GPUs.
• Explore LatentMoE for faster inference in large models.
• Consider OpenMDW for licensing model weights and data.

Topics

• Open Models
• Model Ecosystem
• Mixture-of-Experts
• Model Licensing
• NVIDIA Nemotron
• Cohere Command A+

Best for: CTO, VP of Engineering/Data, AI Architect, AI Scientist, Machine Learning Engineer, Director of AI/ML

Related on AIssential

• Latest open artifacts (#17): NVIDIA, Arcee, Minimax, DeepSeek, Z.ai and others close an eventful year on a high note — The open model ecosystem is rapidly evolving with diverse architectures, data transparency, and specialized performance.
• [AINews] Good Friday — Gemma 4's open-source release and broad ecosystem support enable powerful, efficient multimodal AI on diverse hardware.
• A Cartography of Open Collaboration in Open Source AI: Mapping Practices, Motivations, and Governance in 14 Open Large Language Model Projects — Open collaboration in LLM development is multifaceted, driven by diverse motivations and structured by varied governance models.
• 2025 Open Models Year in Review — Open models significantly advanced in 2025, rivaling closed-source performance and fostering a more open ecosystem.
• 😸 Anthropic’s $30B War Chest, the $200M Political Battle, and the Flood of Free Chinese Models — The AI landscape is marked by massive investments, political influence, and rapid advancements in both open-source and proprietary models.
• Top AI GitHub Repositories in 2026 — The open-source AI ecosystem is rapidly maturing, driven by local AI, agentic capabilities, and powerful open models.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Interconnects AI.