Kimi K2
Kimi K2 is a state-of-the-art open source large language model series built on a mixture-of-experts (MoE) architecture, featuring 1 trillion total parameters and 32 billion activated parameters for task-specific efficiency. Trained with the Muon optimizer on over 15.5 trillion tokens and stabilized by MuonClip’s attention-logit clamping, it delivers exceptional performance in frontier knowledge, reasoning, mathematics, coding, and general agentic workflows. Moonshot AI provides two variants, Kimi-K2-Base for research-level fine-tuning and Kimi-K2-Instruct pre-trained for immediate chat and tool-driven interactions, enabling both custom development and drop-in agentic capabilities. Benchmarks show it outperforms leading open source peers and rivals top proprietary models in coding tasks and complex task breakdowns, while its 128 K-token context length, tool-calling API compatibility, and support for industry-standard inference engines.
Learn more
GigaChat
GigaChat knows how to answer user questions, maintain a dialogue, write program code, create texts and pictures based on descriptions within a single context. Unlike a foreign neural network, the GigaChat service initially already supports multimodal interaction and communicates more competently in Russian.
The architecture of the GigaChat service is based on the neural network ensemble of the NeONKA (NEural Omnimodal Network with Knowledge-Awareness) model, which includes various neural network models and the method of supervised fine-tuning, reinforcement learning with human feedback. Thanks to this, Sber's new neural network can solve many intellectual tasks: keep up a conversation, write texts, answer factual questions. And the inclusion of the Kandinsky 2.1 model in the ensemble gives the neural network the skill of creating images.
Learn more
DeepSeek-V2
DeepSeek-V2 is a state-of-the-art Mixture-of-Experts (MoE) language model introduced by DeepSeek-AI, characterized by its economical training and efficient inference capabilities. With a total of 236 billion parameters, of which only 21 billion are active per token, it supports a context length of up to 128K tokens. DeepSeek-V2 employs innovative architectures like Multi-head Latent Attention (MLA) for efficient inference by compressing the Key-Value (KV) cache and DeepSeekMoE for cost-effective training through sparse computation. This model significantly outperforms its predecessor, DeepSeek 67B, by saving 42.5% in training costs, reducing the KV cache by 93.3%, and enhancing generation throughput by 5.76 times. Pretrained on an 8.1 trillion token corpus, DeepSeek-V2 excels in language understanding, coding, and reasoning tasks, making it a top-tier performer among open-source models.
Learn more
Nemotron 3 Super
Nemotron-3 Super is part of NVIDIA’s Nemotron 3 family of open models designed to enable advanced agentic AI systems that can reason, plan, and execute multi-step workflows across complex environments. The model introduces a hybrid Mamba-Transformer Mixture-of-Experts architecture that combines the efficiency of state-space Mamba layers with the contextual understanding of transformer attention, allowing it to process long sequences and complex reasoning tasks with high accuracy and throughput. This architecture activates only a subset of model parameters for each token, improving computational efficiency while maintaining strong reasoning capabilities and enabling scalable inference for large workloads. Nemotron-3 Super contains roughly 120 billion parameters with around 12 billion active during inference, accelerating multi-step reasoning and collaborative agent interactions across large contexts.
Learn more
