• Models

Category: Models

Google Gemma 3

Google Gemma 3

Google's Gemma 3 models are making significant strides in the open-source AI community. These models, ranging from 270M to 27B parameters, are designed for high efficiency and versatility, capable of running on devices from smartphones to consumer GPUs. Key features include multimodal capabilities, an expanded context window, multilingual support, and efficient fine-tuning. The models offer high performance with a smaller footprint, making them ideal for on-device applications. With open-source availability, developers can use, modify, and distribute the models for commercial purposes. Safety measures ensure responsible usage, and Gemma 3's accessibility democratizes AI development, enabling innovative and cost-effective solutions.

2 minutes reading | 31 Aug 2025
Bagel AI
Multi-Modal

Bagel AI

In May 2025, ByteDance introduced BAGEL, an open-source multimodal AI model with 7 billion active parameters that excels in text understanding, image generation, video processing, and reasoning, outperforming leading open-source models. BAGEL uses a unified, decoder-only architecture with a Mixture-of-Transformer-Experts (MoT) and dual encoders, making it efficient across diverse modalities. It is trained on a large dataset of interleaved multimodal tokens and is available under the Apache 2.0 license. BAGEL surpasses competitors in benchmarks for multimodal tasks and is praised for its performance and accessibility. It holds potential for applications in creative industries, robotics, and research. Despite facing challenges like dependency requirements, BAGEL is set to drive innovation in AI. Explore its capabilities on GitHub or Hugging Face.

4 minutes reading | 28 May 2025
Cogito v1
Conversational AI

Cogito v1

DeepCogito, based in San Francisco, has introduced the Cogito v1 Preview series of open-source AI models, available in various parameter sizes from 3B to 70B. These models are designed for diverse tasks, from lightweight to heavy-duty challenges, and are freely accessible for commercial use on platforms like Hugging Face and Ollama. They claim to outperform competitors like LLaMA and Qwen, though specific benchmark scores are not disclosed. The models feature hybrid reasoning modes, enabling them to switch between standard and reasoning tasks, and are optimized for coding, STEM tasks, and agentic applications. Trained on over 30 languages with a 128k context length, they are versatile and globally applicable. While these models are early versions, larger ones are anticipated. DeepCogito encourages innovators to explore and utilize these models to unlock AI's potential and drive future advancements.

5 minutes reading | 08 Apr 2025