Blog

NVIDIA releases Nemotron-4 340B, a 340B parameter open-source model optimized for synthetic data generation with a permissive enterprise license.

Alibaba Cloud unleashes Qwen2, a 72B parameter powerhouse under Apache 2.0, rivaling top closed models with superior multilingual capabilities.

Zhipu AI launches GLM-4, a 9B parameter model with 128K context, challenging Llama 3 8B in the open-source arena with multilingual support.

Mistral AI launches Codestral, a specialized 22B parameter model designed for high-performance coding tasks with 32K context and open-source availability.

ByteDance releases Doubao Seed 1.5, a versatile open-source model supporting 50+ scenarios via API and App.

OpenAI releases GPT-4o, a native multimodal model offering real-time voice, vision, and text processing at 50% lower cost than GPT-4 Turbo.

DeepSeek AI unveils DeepSeek V2, a massive 236B parameter MoE model with 21B active parameters, challenging top-tier closed models with open weights and high efficiency.

Snowflake releases Arctic, a 480B MoE model optimized for enterprise workloads with Apache 2.0 licensing. Designed for SQL generation and complex reasoning.

Microsoft releases Phi-3, a 14B parameter open-source model rivaling larger competitors with phone-capable efficiency.