AI21 Launches the World's First Production-Grade Mamba Model: Jamba, Supporting 256K Context Length

Tencent Releases the Official Version of HunYuan-T1 Large Language Model with Significantly Enhanced Reasoning Capabilities

Tencent recently released the official version of its HunYuan large language model series – HunYuan-T1. This new model, built upon the medium-scale HunYuan base model and extensively fine-tuned, demonstrates significantly improved reasoning capabilities, particularly excelling in deep thinking and complex problem-solving. Since the launch of the HunYuan T1-Preview in February, users have experienced faster and more insightful processing. The official release marks a significant upgrade to the product line. The HunYuan-T1 development team leveraged the latest Turbo...

AI Daily: DeepSeek R2 Potentially Launching March 17th; Tencent Releases Hunyuan-TurboS; Pika Adds Video Exchange Functionality

Welcome to the 【AI Daily】column! Your daily guide to exploring the world of Artificial Intelligence. We present daily highlights from the AI field, focusing on developers, helping you understand technology trends and learn about innovative AI product applications. Discover new AI products here: https://top.aibase.com/ 1. Tencent Releases Hunyuan-TurboS: The first ultra-large hybrid Transformer-MambaMoE model makes its debut. Tencent today launched Hunyuan-TurboS on X platform...

Tencent Unveils Hunyuan-TurboS: The First Ultra-Large Hybrid Transformer-Mamba MoE Model

AI21 Labs Releases Jamba 1.6: Breaking Limits in Long-Form Text Processing, Supporting Multiple Languages

AI21 Labs has announced the release of Jamba 1.6, a significant upgrade to its large language model that enhances its capabilities in handling long-form text and expands its support to multiple languages.

Llama3 Transforms into Mamba in Just 3 Days! Inference Speed Improved by 1.5 Times

The Mamba team's research focuses on distilling the large transformer model Llama into Mamba by designing a novel inference decoding algorithm that significantly enhances inference speed. The research aims to leverage Llama's extensive knowledge while reducing the high cost of training large models from scratch. The research team successfully transformed Zephyr-7B and Llama-38B into a linear RNN by combining progressive distillation, supervised fine-tuning, and preference optimization methods.

Codestral Mamba: Faster and Doubles GPT-4's Text Processing Length

Recently, the French AI startup Mistral unveiled a brand-new coding model called Codestral Mamba. This model is not only fast but also capable of handling longer code, helping programmers and developers enhance their work efficiency. Mistral has already gained a reputation in the open-source AI field, and the introduction of Codestral Mamba is particularly eye-catching.Product Entry: https://top.aibase.com/tool/codestral-mamba Codestral Mamba is based on a new architecture called "Mamba," which

ZhiYuan Introduces Vision Mamba for Efficient Visual Task Processing, Reducing Memory Usage by 87%

Vision Mamba incorporates state-space models for visual modeling, enhancing performance in ImageNet classification, COCO object detection, and ADE20k semantic segmentation tasks. The Mamba block integrates bidirectional sequence modeling and positional embeddings for efficient visual task handling. Vision Mamba exhibits linear scalability advantages in high-resolution images and long-sequence multimodal applications. Benchmark tests on the ImageNet-1K dataset demonstrate superior performance for Vision Mamba.

Mamba's New Architecture Challenges Transformer: Performance Surpasses All

The new architecture Mamba challenges the Transformer, being comparable to or surpassing it in language modeling. Mamba achieves linear scaling and improves inference throughput by 5 times, widely applicable in fields such as language, audio, and genomics. The introduction of the Selective State Space Model (S4) is one of Mamba's innovations and has achieved significant success. Hardware-aware algorithms address technical challenges in model computation, and the model code and pre-trained checkpoints have been open-sourced.

AI21 Completes Series C Funding, Challenging OpenAI's Position in the Generative AI Field

AI21 has announced the completion of its Series C funding, raising $53 million, with a total valuation of $1.4 billion and a total funding amount of $208 million. Amid internal turmoil at OpenAI, AI21 has chosen to announce its funding to compete for market share in generative AI. AI21 emphasizes close collaboration with enterprises, focusing on customization, robustness, and security, offering tailored solutions that differ from OpenAI. AI21 has performed well in the competition, welcoming investments from Intel Capital and Comcast Ventures.

Israeli Startup AI21 Labs Secures $155 Million in Funding with Participation from Google and Nvidia

Israeli startup AI21 Labs announced the completion of $155 million in funding, with Google and Nvidia also participating. AI21 Labs focuses on customized generative AI models, boasting its own large model, Jurassic-2, competing with OpenAI to provide more accurate and reliable text generation services.