Teaming up with
excellent open-source foundation models

Text

• M

  Meta-Llama-3.1-405B-Instruct

  Llama-3.1-405B-Instruct is a large language model featuring 405 billion parameters, instruction-tuned for enhanced task performance. Built on Meta's Llama architecture, it excels at reasoning, code generation, analysis, and multi-turn conversations, offers strong instruction-following capabilities, consistent output quality, and comprehensive contextual understanding.

• L

  Llama-3.1-Nemotron-70B-Instruct

  Llama-3.1-Nemotron-70B-Instruct is a large language model customized by NVIDIA to improve the helpfulness of LLM generated responses to user queries. As of 1 Oct 2024, this model is #1 on all three automatic alignment benchmarks (verified tab for AlpacaEval 2 LC), edging out strong frontier models such as GPT-4o and Claude 3.5 Sonnet.

• M

  Meta-Llama-3.1-70B-Instruct

  8K

  A collection of pretrained and instruction tuned generative text models in 8 and 70B sizes.

• M

  Meta-Llama-3.1-8B-Instruct

  8K

  A collection of pretrained and instruction tuned generative text models in 8 and 70B sizes.

• L

  Llama-3.2-3B-Instruct

  32K

  The lightweight 3B models are highly capable with multilingual text generation and tool calling abilities. These models empower developers to build personalized, on-device agentic applications with strong privacy where data never leaves the device.

• L

  Llama-3.2-1B-Instruct

  32K

  The lightweight 1B models are highly capable with multilingual text generation and tool calling abilities. These models empower developers to build personalized, on-device agentic applications with strong privacy where data never leaves the device.

• G

  gemma-2-27b-it

  8K

  Gemma is a state-of-the-art, lightweight, open English text model suite from Google.

• G

  gemma-2-9b-it

  8K

  Gemma is a state-of-the-art, lightweight, open English text model suite from Google.

• Q

  Qwen2.5-72B-Instruct

  Qwen2.5 is an advanced language model series offering improved coding, mathematics, instruction-following, and multilingual support (29+ languages).

• Q

  Qwen2.5-Coder-32B-Instruct

  32K

  Qwen2.5-Coder-32B-Instruct, trained on 5.5 trillion tokens, excels in code generation, reasoning, and fixing. Notably, its 32B model matches GPT-4o’s coding capabilities.

• Q

  Qwen2.5-7B-Instruct

  Qwen2.5 is an advanced language model series offering improved coding, mathematics, instruction-following, and multilingual support (29+ languages).

• D

  DeepSeek-V2.5

  DeepSeek-V2.5 is an advanced language model by DeepSeek-AI, featuring enhanced reasoning, coding, and analytical capabilities. It builds upon previous versions with improved context understanding and task completion accuracy. The model excels at technical tasks, mathematical problem-solving, and maintaining high-quality output across diverse applications and domains.

• M

  MythoMax-L2-13b

  4K

  An improved, potentially even perfected variant of MythoMix, MythoMax-L2-13b is proficient at both roleplaying and storywriting due to its unique nature.