| OpenAI o1-mini | OpenAI | A faster, more cost-effective version of o1, optimized for efficiency. |
| OpenAI o1 | OpenAI | Designed for complex, multi-step tasks with advanced accuracy. |
| OpenAI GPT-4o mini | OpenAI | A smaller, more affordable version of GPT-4o, suitable for enterprises and developers. |
| OpenAI GPT-4o | OpenAI | Capable of processing and generating text, images, and audio, offering enhanced functionalities. |
| Claude 3.5 | Anthropic | An advanced version of Claude with improved conversation capabilities and reasoning power. |
| Claude 3.5 Opus | Anthropic | A more capable version of Claude, handling larger and more complex queries. |
| Claude 3 | Anthropic | An advanced conversational AI model with improved performance across various tasks. |
| Claude 3.5 Sonnet | Anthropic | A high-performance model from the Claude family, optimized for conversational tasks. |
| Google Gemini 1.5 Pro | Google | Google's latest conversational AI model, integrated into various applications for enhanced interactions. |
| Google Gemini 1.5 Flash | Google | A faster, more efficient version of Gemini 1.5 Pro, optimized for quick responses. |
| meta-llama/Meta-Llama-3.1-405B-Instruct | Meta | Large-scale instruction-tuned model for advanced generative tasks. |
| meta-llama/Meta-Llama-3.1-8B-Instruct-Turbo | Meta | Optimized Meta Llama 3.1 family of models, instruction-tuned and available in 8B, 70B, and 405B sizes. |
| meta-llama/Llama-3.2-3B-Instruct | Meta | Multilingual Llama 3.2 model available in 1B and 3B sizes, pretrained and instruction-tuned. |
| meta-llama/Llama-3.2-1B-Instruct | Meta | Multilingual Llama 3.2 model available in 1B and 3B sizes, pretrained and instruction-tuned. |
| meta-llama/Llama-3.3-70B-Instruct | Meta | Multilingual LLM trained on 15 trillion tokens, fine-tuned for instruction-following and dialogue. |
| meta-llama/Llama-3.2-11B-Vision-Instruct | Meta | Multimodal model specializing in visual and textual integration for image-based AI tasks. |
| meta-llama/Meta-Llama-3.1-8B-Instruct | Meta | Smaller 8B parameter version of the Llama 3.1 series, optimized for text generation tasks. |
| meta-llama/Meta-Llama-3.1-70B-Instruct-Turbo | Meta | Turbo-optimized version of Meta's Llama 3.1 70B, designed for faster performance with similar pricing. |
| meta-llama/Meta-Llama-3.1-70B-Instruct | Meta | Pretrained and instruction-tuned for various generative text tasks, available in sizes from 8B to 405B. |
| meta-llama/Llama-3.3-70B-Instruct-Turbo | Meta | Optimized version of Llama 3.3-70B, using FP8 quantization for faster inference with minor accuracy trade-off. |
| meta-llama/Llama-3.2-90B-Vision-Instruct | Qwen | Multimodal model excelling in visual tasks like image captioning and question answering. |
| Qwen/Qwen2.5-Coder-7B | Qwen | Code-specific large language model designed for code generation, reasoning, and fixing tasks. |
| Qwen/Qwen2-7B-Instruct | Qwen | Excels in language understanding, multilingual capabilities, coding, mathematics, and reasoning. |
| Qwen/Qwen2.5-Coder-32B-Instruct | Qwen | Code-specific model designed for code generation, reasoning, and fixing. |
| Qwen/Qwen2.5-72B-Instruct | Qwen | Pretrained on up to 18 trillion tokens, offering improvements in knowledge, coding, and instruction following. |
| Qwen/Qwen2.5-7B-Instruct | Qwen | Focuses on language understanding, multilingual capabilities, coding, and reasoning, but is now deprecated. |
| Qwen/Qwen2-72B-Instruct | Qwen | Excels in language understanding, multilingual capabilities, coding, and reasoning with a 72 billion parameter model. |
| Qwen/QwQ-32B-Preview | NVIDIA | Experimental model focusing on AI reasoning with impressive scores in analytical tasks. |
| nvidia/Llama-3.1-Nemotron-70B-Instruct | Microsoft | Customized Llama-3.1 variant optimized for improved user response helpfulness. |
| microsoft/phi-4 | Microsoft | Phi-4 is built for high-quality and advanced reasoning with datasets from public domain websites, academic books, and Q&A datasets. |
| microsoft/WizardLM-2-8x22B | 01-ai | Advanced model highly competitive with leading proprietary models. |
| 01-ai/Yi-34B-Chat | Gryphe | Large model designed for dynamic conversational tasks, although replaced. |
| Gryphe/MythoMax-L2-13b | Gryphe | Creative model designed for generating imaginative and coherent narratives. |
| Gryphe/MythoMax-L2-13b-turbo | HuggingFace | Faster version of MythoMax-L2-13b, running on multiple H100 cards in fp8 precision for higher throughput. |
| HuggingFaceH4/zephyr-orpo-141b-A35b-v0.1 | KoboldAI | Instruction-tuned version of Mixtral-8x22B, offering strong performance in chat benchmarks. |
| KoboldAI/LLaMA2-13B-Tiefighter | NousResearch | Creative model fine-tuned for storytelling and adventure, generating engaging content. |
| NousResearch/Hermes-3-Llama-3.1-405B | Phind | Fine-tuned version of Llama-3.1 405B, designed for roleplaying, reasoning, conversation, and improved code generation. |
| Phind/Phind-CodeLlama-34B-v2 | Sao10K | Fine-tuned for programming-related tasks with a focus on Python, C/C++, TypeScript, and Java. |
| Sao10K/L3-8B-Lunaris-v1 | Sao10K | Generalist/roleplaying model optimized for a wide range of creative tasks. |
| Sao10K/L3.1-70B-Euryale-v2.2 | Sao10K | Focused on creative roleplay, offering enhanced narrative capabilities. |
| Sao10K/L3-70B-Euryale-v2.1 | bigcode | Focuses on creative roleplay tasks, designed for imaginative storytelling and interactive use. |
| bigcode/starcoder2-15b-instruct-v0.1 | bigcode | Self-aligned code LLM trained with a fully permissive and transparent pipeline. |
| bigcode/starcoder2-15b | codellama | Trained on 600+ programming languages, specializing in code completion tasks. |
| codellama/CodeLlama-34b-Instruct-hf | codellama | Advanced model specializing in code generation and understanding both code and natural language prompts. |
| codellama/CodeLlama-70b-Instruct-hf | cognitivecomputations | Largest and latest code generation model from the Code Llama collection. |
| cognitivecomputations/dolphin-2.6-mixtral-8x7b | cognitivecomputations | Fine-tuned version of Mixtral-8x7b for coding tasks, uncensored and with a compliance focus. |
| cognitivecomputations/dolphin-2.9.1-llama-3-70b | deepseek-ai | Fine-tuned Llama-3-70b model, focused on improving compliance, instruction, and function calling abilities. |
| deepseek-ai/DeepSeek-V3 | mistralai | A cutting-edge large language model designed for advanced reasoning, conversation, and comprehension tasks. |
| mistralai/Mistral-7B-Instruct-v0.3 | mistralai | Mistral-7B-Instruct-v0.3 is instruction-tuned, with larger vocabulary, newer tokenizer, and supports function calling. |
| mistralai/Mistral-Nemo-Instruct-2407 | Google | 12B model trained by Mistral AI and NVIDIA, significantly outperforming existing models of similar size. |
| google/gemma-1.1-7b-it | Google | Gemma 1.1 7B is instruction-tuned with novel RLHF method, enhancing coding capabilities and factuality. |
| google/gemma-2-27b-it | Google | Gemma-2-27B offers competitive performance against larger models, excelling in multi-turn conversations. |
| google/codegemma-7b-it | Google | Lightweight code models built on top of Gemma, specializing in code completion and generation tasks. |
| google/gemma-2-9b-it | Unknown | Gemma-2 9B outperforms Llama 3 8B and other open models in its size category. |
