Meta's flagship open-weights model, optimized for complex instruction following, reasoning, and tool use.

Lightweight 8B Meta Llama 3 model for fast, efficient chat tasks.

DeepSeek's latest V4 Pro model with strong tool use and general reasoning capabilities.

Massive 671B DeepCogito architecture for advanced cognitive processing and deduction.

Colossal 480B MoE coding model for ultimate software engineering and refactoring intelligence.

Qwen3.5 397B — massive parameter model for complex reasoning and generation.

High-throughput 235B Qwen3 model with tool support, optimized for scale.

Compact and fast Qwen3.5 model for efficient inference and standard chat.

Fast 7B Qwen model with excellent instruction following, tool use, and low latency.

Moonshot AI Kimi K2.6 — strong agentic model with advanced tool use capabilities.

MiniMax M2.7 — high-performance foundation model for conversational and creative tasks.

Zhipu AI GLM-5 — next-generation foundation model with extended context and reasoning.

Zhipu AI GLM-5.1 — improved iteration with enhanced multilingual and reasoning capabilities.

OpenAI's 20B open-source model for standard conversational and instruction-following tasks.

OpenAI's 120B open-source model for complex tasks requiring large parameter capacity.

Google's lightweight Gemma architecture fine-tuned for instruction following and tool execution.