Run Any LLM Without
Out-Of-Memory Errors.

Mistral AI's specialized open-weights code generation model supporting over 80+ programming languages.

Cohere's enterprise-grade model built explicitly for highly advanced Retrieval-Augmented Generation (RAG) tasks and automated agents.

DeepSeek's 2026 flagship Mixture-of-Experts (MoE) titan. It introduces a massive 1-million-token context window with deep architectural optimizations, offering frontier-level intelligence at an unprecedented 75% price reduction.

An efficiency-optimized Mixture-of-Experts (MoE) variant featuring 284B total and 13B active parameters, specifically designed for high-throughput and ultra-low latency while maintaining a 1M context length.

An upgraded iteration of DeepSeek's v3 line architecture, optimizing mathematical execution and multi-turn enterprise workflows to peak efficiency.

A Mixture-of-Experts (MoE) monster from DeepSeek with 671B total and 37B active parameters, delivering world-class intelligence with extreme architectural efficiency.

DeepSeek's flagship reasoning model utilizing advanced reinforcement learning. Specializes in chain-of-thought processing for complex math and coding.

Google's premier dense open model from the April 2026 generation. Released under the permissive Apache 2.0 license, it incorporates Gemini 3 architecture to deliver cross-modal reasoning and advanced agentic multi-step planning.

Google's highly flexible Mixture-of-Experts model utilizing a sparse architecture to deliver 31B-class text and vision performance at a fraction of the inference cost.

Google's breakthrough edge-tier model that natively processes text, vision, and real-time audio waveforms, designed to bring fully autonomous multimodal agents to local consumer hardware.

Google's lightweight star, packing punchy, forward-thinking architecture that outperforms many models twice its size.

Google's highly advanced mid-sized model utilizing an innovative sliding window attention mechanism for high-efficiency processing.

Zhipu AI's (Z.ai) massive open-weights titan designed for complex systems engineering. Features 40B active parameters per token, a 200K input pipeline, and unprecedented multi-step agentic execution scoring.

OpenAI's historical, highly disruptive release into the open-weights community. Built with an Apache 2.0 license, it brings OpenAI's advanced reasoning behaviors straight to self-hosted cloud environments.

Shanghai AI Lab's flagship model, outstanding in long-context processing with perfect information retrieval up to 128K tokens.

Moonshot AI's 1-trillion parameter Mixture-of-Experts (MoE) powerhouse (32B active). It stands at the absolute frontier of open-weights models, specifically dominating complex front-end visual coding and multi-agent systems.

Meta's next-generation sparse Mixture-of-Experts (MoE) model. Activating 17B parameters per token, it brings a massive 10-million-token context window alongside native multimodal text-and-image understanding.

Meta's highly optimized lightweight model with an expanded 128K context window, perfect for edge deployment and efficient fine-tuning.

Meta's flagship 70B model, renowned for elite tier reasoning, coding, and general knowledge capabilities.

An open-weight productivity power-house utilizing a sparse MoE layout (10B active parameters). It excels in real-world software engineering workflows and long-form context synthesis up to 196K tokens.

Mistral AI's highly versatile 'all-in-one' model that merges advanced reasoning, vision (Pixtral), and agentic coding into a single highly optimized sparse MoE framework.

Collaborative model built by Mistral AI and NVIDIA, featuring a 128K context window and top-tier reasoning in a compact size.

Mistral's leading frontier model designed for complex multilingual enterprise tasks, coding, and function orchestration.

NVIDIA's specialized mid-tier model designed specifically to optimize guardrails, RAG alignment, and prompt engineering orchestration.

Microsoft's heavily requested compact reasoning engine. Fine-tuned specifically from the Phi-4 base to emulate advanced math and scientific data deduction without the resource overhead.

Microsoft's heavy-hitting small language model, trained on heavily curated high-quality synthetic data and web texts.

Alibaba's premier open-source sparse MoE model (22B active parameters). Released under the Apache 2.0 license, it sets the enterprise benchmark for multilingual processing, advanced tool use, and visual reasoning.

Alibaba's efficient model packed with advanced coding skills and industry-leading multilingual capabilities.

Alibaba's top-tier open weights titan, matching closed-source models in mathematics, multi-turn coding, and multilingual knowledge.

Hugging Face's meticulously trained ultra-compact model, engineered from curated high-quality datasets to provide unmatched instruction-following logic on local edge hardware.

01.AI's highly optimized dense model built upon the acclaimed Yi architecture, offering enhanced coding, math, and reasoning capabilities for enterprise applications.

01.AI's elite Mixture-of-Experts engine, delivering outstanding cross-disciplinary academic and conversational responses.