Mistral

Llama is Meta's family of open-weight large language models that became the de-facto foundation for the open-source AI ecosystem. Available in sizes from a few billion to hundreds of billions of parameters, Llama models can be fine-tuned and self-hosted with a permissive licence. They power countless products via providers like Groq, Together and Fireworks.

DeepSeek is a Chinese AI lab that stunned the industry with frontier-level reasoning models at a fraction of typical costs. Its R-series reasoning models and V-series chat models are open-weight and extremely cheap via API. DeepSeek is the go-to choice for developers who need strong math, coding and reasoning performance on a tight budget.

Qwen is Alibaba's series of open-weight models spanning chat, coding, vision and math. The lineup is among the strongest open releases, with excellent multilingual ability and competitive benchmark scores. Qwen models are widely used across Asia and increasingly worldwide for self-hosted and API-based deployments.

Claude is Anthropic's family of AI assistants, known for long-context reasoning, careful writing and strong coding ability. The Opus, Sonnet and Haiku tiers let you trade off intelligence, speed and cost. Claude excels at nuanced analysis, document understanding and agentic coding workflows, and is a favourite among developers and writers who value clarity.