From 540176059ab7a9bcc7e83871da04c60fff8b2ecb Mon Sep 17 00:00:00 2001 From: Enrico Ros Date: Thu, 30 Jan 2025 14:57:18 -0800 Subject: [PATCH] Add Mistral-3 (24B) --- .../llms/server/ollama/ollama.models.ts | 178 +++++++++--------- 1 file changed, 89 insertions(+), 89 deletions(-) diff --git a/src/modules/llms/server/ollama/ollama.models.ts b/src/modules/llms/server/ollama/ollama.models.ts index 57562b121..18141fa7c 100644 --- a/src/modules/llms/server/ollama/ollama.models.ts +++ b/src/modules/llms/server/ollama/ollama.models.ts @@ -15,12 +15,12 @@ * Note: the default contextWindow in code is 8192, so we do not redefine that */ export const OLLAMA_BASE_MODELS: { [key: string]: { description: string, pulls: number, contextWindow?: number, hasTools?: true, hasVision?: true, isEmbeddings?: true, tags?: string[], added?: string } } = { - 'deepseek-r1': { description: 'DeepSeek\'s first-generation of reasoning models with comparable performance to OpenAI-o1, including six dense models distilled from DeepSeek-R1 based on Llama and Qwen.', pulls: 3500000, tags: [ '1.5b', '7b', '8b', '14b', '32b', '70b', '671b' ], added: '20250128' }, - 'llama3.3': { description: 'New state of the art 70B model. Llama 3.3 70B offers similar performance compared to the Llama 3.1 405B model.', pulls: 923400, tags: [ '70b' ], hasTools: true, added: '20241210' }, - 'phi4': { description: 'Phi-4 is a 14B parameter, state-of-the-art open model from Microsoft.', pulls: 214700, tags: [ '14b' ], added: '20250128' }, - 'llama3.2': { description: 'Meta\'s Llama 3.2 goes small with 1B and 3B models. ', pulls: 7700000, tags: [ '1b', '3b' ], hasTools: true, added: '20241210' }, - 'llama3.1': { description: 'Llama 3.1 is a new state-of-the-art model from Meta available in 8B, 70B and 405B parameter sizes.', pulls: 20800000, tags: [ '8b', '70b', '405b' ], hasTools: true, added: '20241210' }, - 'nomic-embed-text': { description: 'A high-performing open embedding model with a large token context window.', pulls: 13100000, tags: [], isEmbeddings: true, added: '20240501' }, + 'deepseek-r1': { description: 'DeepSeek\'s first-generation of reasoning models with comparable performance to OpenAI-o1, including six dense models distilled from DeepSeek-R1 based on Llama and Qwen.', pulls: 4200000, tags: [ '1.5b', '7b', '8b', '14b', '32b', '70b', '671b' ], added: '20250128' }, + 'llama3.3': { description: 'New state of the art 70B model. Llama 3.3 70B offers similar performance compared to the Llama 3.1 405B model.', pulls: 980600, tags: [ '70b' ], hasTools: true, added: '20241210' }, + 'phi4': { description: 'Phi-4 is a 14B parameter, state-of-the-art open model from Microsoft.', pulls: 225700, tags: [ '14b' ], added: '20250128' }, + 'llama3.2': { description: 'Meta\'s Llama 3.2 goes small with 1B and 3B models. ', pulls: 7800000, tags: [ '1b', '3b' ], hasTools: true, added: '20241210' }, + 'llama3.1': { description: 'Llama 3.1 is a new state-of-the-art model from Meta available in 8B, 70B and 405B parameter sizes.', pulls: 21000000, tags: [ '8b', '70b', '405b' ], hasTools: true, added: '20241210' }, + 'nomic-embed-text': { description: 'A high-performing open embedding model with a large token context window.', pulls: 13200000, tags: [], isEmbeddings: true, added: '20240501' }, 'mistral': { description: 'The 7B model released by Mistral AI, updated to version 0.3.', pulls: 8300000, tags: [ '7b' ], hasTools: true }, 'llama3': { description: 'Meta Llama 3: The most capable openly available LLM to date', pulls: 7300000, tags: [ '8b', '70b' ], added: '20240501' }, 'qwen': { description: 'Qwen 1.5 is a series of large language models by Alibaba Cloud spanning from 0.5B to 110B parameters', pulls: 4300000, tags: [ '0.5b', '1.8b', '4b', '7b', '14b', '32b', '72b', '110b' ] }, @@ -29,108 +29,108 @@ export const OLLAMA_BASE_MODELS: { [key: string]: { description: string, pulls: 'qwen2.5': { description: 'Qwen2.5 models are pretrained on Alibaba\'s latest large-scale dataset, encompassing up to 18 trillion tokens. The model supports up to 128K tokens and has multilingual support. ', pulls: 3400000, tags: [ '0.5b', '1.5b', '3b', '7b', '14b', '32b', '72b' ], hasTools: true, added: '20241210' }, 'phi3': { description: 'Phi-3 is a family of lightweight 3B (Mini) and 14B (Medium) state-of-the-art open models by Microsoft.', pulls: 2800000, tags: [ '3.8b', '14b' ], added: '20240501' }, 'llama2': { description: 'Llama 2 is a collection of foundation language models ranging from 7B to 70B parameters.', pulls: 2800000, tags: [ '7b', '13b', '70b' ] }, - 'llava': { description: '๐ŸŒ‹ LLaVA is a novel end-to-end trained large multimodal model that combines a vision encoder and Vicuna for general-purpose visual and language understanding. Updated to version 1.6.', pulls: 2700000, tags: [ '7b', '13b', '34b' ], hasVision: true }, + 'llava': { description: '๐ŸŒ‹ LLaVA is a novel end-to-end trained large multimodal model that combines a vision encoder and Vicuna for general-purpose visual and language understanding. Updated to version 1.6.', pulls: 2800000, tags: [ '7b', '13b', '34b' ], hasVision: true }, 'gemma2': { description: 'Google Gemma 2 is a high-performing and efficient model available in three sizes: 2B, 9B, and 27B.', pulls: 2500000, tags: [ '2b', '9b', '27b' ], added: '20240628' }, 'qwen2.5-coder': { description: 'The latest series of Code-Specific Qwen models, with significant improvements in code generation, code reasoning, and code fixing.', pulls: 1700000, tags: [ '0.5b', '1.5b', '3b', '7b', '14b', '32b' ], hasTools: true, added: '20241210' }, 'codellama': { description: 'A large language model that can use text prompts to generate and discuss code.', pulls: 1700000, tags: [ '7b', '13b', '34b', '70b' ] }, 'tinyllama': { description: 'The TinyLlama project is an open endeavor to train a compact 1.1B Llama model on 3 trillion tokens.', pulls: 1200000, tags: [ '1.1b' ] }, - 'mxbai-embed-large': { description: 'State-of-the-art large embedding model from mixedbread.ai', pulls: 1100000, tags: [ '335m' ], isEmbeddings: true, added: '20240501' }, + 'mxbai-embed-large': { description: 'State-of-the-art large embedding model from mixedbread.ai', pulls: 1200000, tags: [ '335m' ], isEmbeddings: true, added: '20240501' }, 'mistral-nemo': { description: 'A state-of-the-art 12B model with 128k context length, built by Mistral AI in collaboration with NVIDIA.', pulls: 1000000, tags: [ '12b' ], hasTools: true, added: '20241210' }, - 'llama3.2-vision': { description: 'Llama 3.2 Vision is a collection of instruction-tuned image reasoning generative models in 11B and 90B sizes.', pulls: 964900, tags: [ '11b', '90b' ], hasVision: true, added: '20241210' }, - 'starcoder2': { description: 'StarCoder2 is the next generation of transparently trained open code LLMs that comes in three sizes: 3B, 7B and 15B parameters. ', pulls: 780600, tags: [ '3b', '7b', '15b' ], added: '20240501' }, - 'snowflake-arctic-embed': { description: 'A suite of text embedding models by Snowflake, optimized for performance.', pulls: 561300, tags: [ '22m', '33m', '110m', '137m', '335m' ], isEmbeddings: true, added: '20240501' }, - 'mixtral': { description: 'A set of Mixture of Experts (MoE) model with open weights by Mistral AI in 8x7b and 8x22b parameter sizes.', pulls: 536800, tags: [ '8x7b', '8x22b' ], hasTools: true }, - 'deepseek-coder-v2': { description: 'An open-source Mixture-of-Experts code language model that achieves performance comparable to GPT4-Turbo in code-specific tasks.', pulls: 496400, tags: [ '16b', '236b' ], added: '20240628' }, - 'dolphin-mixtral': { description: 'Uncensored, 8x7b and 8x22b fine-tuned models based on the Mixtral mixture of experts models that excels at coding tasks. Created by Eric Hartford.', pulls: 482900, tags: [ '8x7b', '8x22b' ] }, - 'phi': { description: 'Phi-2: a 2.7B language model by Microsoft Research that demonstrates outstanding reasoning and language understanding capabilities.', pulls: 471800, tags: [ '2.7b' ] }, - 'codegemma': { description: 'CodeGemma is a collection of powerful, lightweight models that can perform a variety of coding tasks like fill-in-the-middle code completion, code generation, natural language understanding, mathematical reasoning, and instruction following.', pulls: 470900, tags: [ '2b', '7b' ], added: '20240501' }, - 'deepseek-coder': { description: 'DeepSeek Coder is a capable coding model trained on two trillion code and natural language tokens.', pulls: 431300, tags: [ '1.3b', '6.7b', '33b' ] }, - 'llama2-uncensored': { description: 'Uncensored Llama 2 model by George Sung and Jarrad Hope.', pulls: 430000, tags: [ '7b', '70b' ] }, - 'wizardlm2': { description: 'State of the art large language model from Microsoft AI with improved performance on complex chat, multilingual, reasoning and agent use cases.', pulls: 322000, tags: [ '7b', '8x22b' ], added: '20240501' }, - 'dolphin-mistral': { description: 'The uncensored Dolphin model based on Mistral that excels at coding tasks. Updated to version 2.8.', pulls: 308500, tags: [ '7b' ] }, - 'all-minilm': { description: 'Embedding models on very large sentence level datasets.', pulls: 276000, tags: [ '22m', '33m' ], isEmbeddings: true, added: '20240501' }, - 'dolphin-llama3': { description: 'Dolphin 2.9 is a new model with 8B and 70B sizes by Eric Hartford based on Llama 3 that has a variety of instruction, conversational, and coding skills.', pulls: 273800, tags: [ '8b', '70b' ], added: '20240501' }, - 'command-r': { description: 'Command R is a Large Language Model optimized for conversational interaction and long context tasks.', pulls: 268300, tags: [ '35b' ], hasTools: true, added: '20240501' }, - 'orca-mini': { description: 'A general-purpose model ranging from 3 billion parameters to 70 billion, suitable for entry-level hardware.', pulls: 266500, tags: [ '3b', '7b', '13b', '70b' ] }, - 'bge-m3': { description: 'BGE-M3 is a new model from BAAI distinguished for its versatility in Multi-Functionality, Multi-Linguality, and Multi-Granularity.', pulls: 265500, tags: [ '567m' ], isEmbeddings: true, added: '20241210' }, - 'yi': { description: 'Yi 1.5 is a high-performing, bilingual language model.', pulls: 259600, tags: [ '6b', '9b', '34b' ] }, - 'llava-llama3': { description: 'A LLaVA model fine-tuned from Llama 3 Instruct with better scores in several benchmarks.', pulls: 252600, tags: [ '8b' ], hasVision: true, added: '20240628' }, - 'zephyr': { description: 'Zephyr is a series of fine-tuned versions of the Mistral and Mixtral models that are trained to act as helpful assistants.', pulls: 232200, tags: [ '7b', '141b' ] }, - 'phi3.5': { description: 'A lightweight AI model with 3.8 billion parameters with performance overtaking similarly and larger sized models.', pulls: 225700, tags: [ '3.8b' ], added: '20241210' }, - 'codestral': { description: 'Codestral is Mistral AI\'s first-ever code model designed for code generation tasks.', pulls: 200200, tags: [ '22b' ], added: '20240628' }, - 'starcoder': { description: 'StarCoder is a code generation model trained on 80+ programming languages.', pulls: 179900, tags: [ '1b', '3b', '7b', '15b' ] }, - 'granite-code': { description: 'A family of open foundation models by IBM for Code Intelligence', pulls: 178200, tags: [ '3b', '8b', '20b', '34b' ], added: '20240628' }, - 'vicuna': { description: 'General use chat model based on Llama and Llama 2 with 2K to 16K context sizes.', pulls: 171500, tags: [ '7b', '13b', '33b' ] }, - 'smollm': { description: '๐Ÿช A family of small models with 135M, 360M, and 1.7B parameters, trained on a new high-quality dataset.', pulls: 165900, tags: [ '135m', '360m', '1.7b' ], added: '20241210' }, + 'llama3.2-vision': { description: 'Llama 3.2 Vision is a collection of instruction-tuned image reasoning generative models in 11B and 90B sizes.', pulls: 971700, tags: [ '11b', '90b' ], hasVision: true, added: '20241210' }, + 'starcoder2': { description: 'StarCoder2 is the next generation of transparently trained open code LLMs that comes in three sizes: 3B, 7B and 15B parameters. ', pulls: 788400, tags: [ '3b', '7b', '15b' ], added: '20240501' }, + 'snowflake-arctic-embed': { description: 'A suite of text embedding models by Snowflake, optimized for performance.', pulls: 568400, tags: [ '22m', '33m', '110m', '137m', '335m' ], isEmbeddings: true, added: '20240501' }, + 'mixtral': { description: 'A set of Mixture of Experts (MoE) model with open weights by Mistral AI in 8x7b and 8x22b parameter sizes.', pulls: 539400, tags: [ '8x7b', '8x22b' ], hasTools: true }, + 'deepseek-coder-v2': { description: 'An open-source Mixture-of-Experts code language model that achieves performance comparable to GPT4-Turbo in code-specific tasks.', pulls: 504000, tags: [ '16b', '236b' ], added: '20240628' }, + 'dolphin-mixtral': { description: 'Uncensored, 8x7b and 8x22b fine-tuned models based on the Mixtral mixture of experts models that excels at coding tasks. Created by Eric Hartford.', pulls: 484100, tags: [ '8x7b', '8x22b' ] }, + 'phi': { description: 'Phi-2: a 2.7B language model by Microsoft Research that demonstrates outstanding reasoning and language understanding capabilities.', pulls: 473200, tags: [ '2.7b' ] }, + 'codegemma': { description: 'CodeGemma is a collection of powerful, lightweight models that can perform a variety of coding tasks like fill-in-the-middle code completion, code generation, natural language understanding, mathematical reasoning, and instruction following.', pulls: 472100, tags: [ '2b', '7b' ], added: '20240501' }, + 'deepseek-coder': { description: 'DeepSeek Coder is a capable coding model trained on two trillion code and natural language tokens.', pulls: 438800, tags: [ '1.3b', '6.7b', '33b' ] }, + 'llama2-uncensored': { description: 'Uncensored Llama 2 model by George Sung and Jarrad Hope.', pulls: 432100, tags: [ '7b', '70b' ] }, + 'wizardlm2': { description: 'State of the art large language model from Microsoft AI with improved performance on complex chat, multilingual, reasoning and agent use cases.', pulls: 325700, tags: [ '7b', '8x22b' ], added: '20240501' }, + 'dolphin-mistral': { description: 'The uncensored Dolphin model based on Mistral that excels at coding tasks. Updated to version 2.8.', pulls: 309200, tags: [ '7b' ] }, + 'all-minilm': { description: 'Embedding models on very large sentence level datasets.', pulls: 276800, tags: [ '22m', '33m' ], isEmbeddings: true, added: '20240501' }, + 'dolphin-llama3': { description: 'Dolphin 2.9 is a new model with 8B and 70B sizes by Eric Hartford based on Llama 3 that has a variety of instruction, conversational, and coding skills.', pulls: 274600, tags: [ '8b', '70b' ], added: '20240501' }, + 'command-r': { description: 'Command R is a Large Language Model optimized for conversational interaction and long context tasks.', pulls: 270000, tags: [ '35b' ], hasTools: true, added: '20240501' }, + 'bge-m3': { description: 'BGE-M3 is a new model from BAAI distinguished for its versatility in Multi-Functionality, Multi-Linguality, and Multi-Granularity.', pulls: 268300, tags: [ '567m' ], isEmbeddings: true, added: '20241210' }, + 'orca-mini': { description: 'A general-purpose model ranging from 3 billion parameters to 70 billion, suitable for entry-level hardware.', pulls: 266700, tags: [ '3b', '7b', '13b', '70b' ] }, + 'yi': { description: 'Yi 1.5 is a high-performing, bilingual language model.', pulls: 259800, tags: [ '6b', '9b', '34b' ] }, + 'llava-llama3': { description: 'A LLaVA model fine-tuned from Llama 3 Instruct with better scores in several benchmarks.', pulls: 253500, tags: [ '8b' ], hasVision: true, added: '20240628' }, + 'zephyr': { description: 'Zephyr is a series of fine-tuned versions of the Mistral and Mixtral models that are trained to act as helpful assistants.', pulls: 232400, tags: [ '7b', '141b' ] }, + 'phi3.5': { description: 'A lightweight AI model with 3.8 billion parameters with performance overtaking similarly and larger sized models.', pulls: 226500, tags: [ '3.8b' ], added: '20241210' }, + 'codestral': { description: 'Codestral is Mistral AI\'s first-ever code model designed for code generation tasks.', pulls: 201300, tags: [ '22b' ], added: '20240628' }, + 'starcoder': { description: 'StarCoder is a code generation model trained on 80+ programming languages.', pulls: 180100, tags: [ '1b', '3b', '7b', '15b' ] }, + 'granite-code': { description: 'A family of open foundation models by IBM for Code Intelligence', pulls: 178500, tags: [ '3b', '8b', '20b', '34b' ], added: '20240628' }, + 'vicuna': { description: 'General use chat model based on Llama and Llama 2 with 2K to 16K context sizes.', pulls: 171600, tags: [ '7b', '13b', '33b' ] }, + 'smollm': { description: '๐Ÿช A family of small models with 135M, 360M, and 1.7B parameters, trained on a new high-quality dataset.', pulls: 166300, tags: [ '135m', '360m', '1.7b' ], added: '20241210' }, + 'wizard-vicuna-uncensored': { description: 'Wizard Vicuna Uncensored is a 7B, 13B, and 30B parameter model based on Llama 2 uncensored by Eric Hartford.', pulls: 164900, tags: [ '7b', '13b', '30b' ] }, 'mistral-openorca': { description: 'Mistral OpenOrca is a 7 billion parameter model, fine-tuned on top of the Mistral 7B model using the OpenOrca dataset.', pulls: 164400, tags: [ '7b' ] }, - 'wizard-vicuna-uncensored': { description: 'Wizard Vicuna Uncensored is a 7B, 13B, and 30B parameter model based on Llama 2 uncensored by Eric Hartford.', pulls: 164300, tags: [ '7b', '13b', '30b' ] }, - 'qwq': { description: 'QwQ is an experimental research model focused on advancing AI reasoning capabilities.', pulls: 151800, tags: [ '32b' ], hasTools: true, added: '20241210' }, + 'qwq': { description: 'QwQ is an experimental research model focused on advancing AI reasoning capabilities.', pulls: 152100, tags: [ '32b' ], hasTools: true, added: '20241210' }, 'llama2-chinese': { description: 'Llama 2 based model fine tuned to improve Chinese dialogue ability.', pulls: 145300, tags: [ '7b', '13b' ] }, - 'smollm2': { description: 'SmolLM2 is a family of compact language models available in three size: 135M, 360M, and 1.7B parameters.', pulls: 135300, tags: [ '135m', '360m', '1.7b' ], hasTools: true, added: '20241210' }, - 'codegeex4': { description: 'A versatile model for AI software development scenarios, including code completion.', pulls: 132900, tags: [ '9b' ], added: '20241210' }, - 'openchat': { description: 'A family of open-source models trained on a wide variety of data, surpassing ChatGPT on various benchmarks. Updated to version 3.5-0106.', pulls: 130800, tags: [ '7b' ] }, - 'aya': { description: 'Aya 23, released by Cohere, is a new family of state-of-the-art, multilingual models that support 23 languages. ', pulls: 129300, tags: [ '8b', '35b' ], added: '20240628' }, - 'codeqwen': { description: 'CodeQwen1.5 is a large language model pretrained on a large amount of code data.', pulls: 125300, tags: [ '7b' ], added: '20240501' }, - 'deepseek-v3': { description: 'A strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token.', pulls: 122300, tags: [ '671b' ], added: '20250128' }, + 'smollm2': { description: 'SmolLM2 is a family of compact language models available in three size: 135M, 360M, and 1.7B parameters.', pulls: 137600, tags: [ '135m', '360m', '1.7b' ], hasTools: true, added: '20241210' }, + 'codegeex4': { description: 'A versatile model for AI software development scenarios, including code completion.', pulls: 133000, tags: [ '9b' ], added: '20241210' }, + 'openchat': { description: 'A family of open-source models trained on a wide variety of data, surpassing ChatGPT on various benchmarks. Updated to version 3.5-0106.', pulls: 131200, tags: [ '7b' ] }, + 'deepseek-v3': { description: 'A strong Mixture-of-Experts (MoE) language model with 671B total parameters with 37B activated for each token.', pulls: 129700, tags: [ '671b' ], added: '20250128' }, + 'aya': { description: 'Aya 23, released by Cohere, is a new family of state-of-the-art, multilingual models that support 23 languages. ', pulls: 129500, tags: [ '8b', '35b' ], added: '20240628' }, + 'codeqwen': { description: 'CodeQwen1.5 is a large language model pretrained on a large amount of code data.', pulls: 125400, tags: [ '7b' ], added: '20240501' }, 'nous-hermes2': { description: 'The powerful family of models by Nous Research that excels at scientific discussion and coding tasks.', pulls: 119800, tags: [ '10.7b', '34b' ] }, - 'mistral-large': { description: 'Mistral Large 2 is Mistral\'s new flagship model that is significantly more capable in code generation, mathematics, and reasoning with 128k context window and support for dozens of languages.', pulls: 117700, tags: [ '123b' ], hasTools: true, added: '20241210' }, - 'command-r-plus': { description: 'Command R+ is a powerful, scalable large language model purpose-built to excel at real-world enterprise use cases.', pulls: 116100, contextWindow: 128000, tags: [ '104b' ], hasTools: true, added: '20240501' }, - 'openhermes': { description: 'OpenHermes 2.5 is a 7B model fine-tuned by Teknium on Mistral with fully open datasets.', pulls: 115700 }, - 'stable-code': { description: 'Stable Code 3B is a coding model with instruct and code completion variants on par with models such as Code Llama 7B that are 2.5x larger.', pulls: 115700, tags: [ '3b' ] }, - 'tinydolphin': { description: 'An experimental 1.1B parameter model trained on the new Dolphin 2.8 dataset by Eric Hartford and based on TinyLlama.', pulls: 114400, tags: [ '1.1b' ] }, - 'glm4': { description: 'A strong multi-lingual general language model with competitive performance to Llama 3.', pulls: 114300, tags: [ '9b' ], added: '20241210' }, - 'wizardcoder': { description: 'State-of-the-art code generation model', pulls: 113600, tags: [ '33b' ] }, - 'qwen2-math': { description: 'Qwen2 Math is a series of specialized math language models built upon the Qwen2 LLMs, which significantly outperforms the mathematical capabilities of open-source models and even closed-source models (e.g., GPT4o).', pulls: 112800, tags: [ '1.5b', '7b', '72b' ], hasTools: true, added: '20241210' }, - 'bakllava': { description: 'BakLLaVA is a multimodal model consisting of the Mistral 7B base model augmented with the LLaVA architecture.', pulls: 104600, tags: [ '7b' ], hasVision: true }, - 'stablelm2': { description: 'Stable LM 2 is a state-of-the-art 1.6B and 12B parameter language model trained on multilingual data in English, Spanish, German, Italian, French, Portuguese, and Dutch.', pulls: 104400, tags: [ '1.6b', '12b' ] }, - 'deepseek-llm': { description: 'An advanced language model crafted with 2 trillion bilingual tokens.', pulls: 101900, tags: [ '7b', '67b' ] }, - 'reflection': { description: 'A high-performing model trained with a new technique called Reflection-tuning that teaches a LLM to detect mistakes in its reasoning and correct course.', pulls: 101700, tags: [ '70b' ], added: '20241210' }, - 'moondream': { description: 'moondream2 is a small vision language model designed to run efficiently on edge devices.', pulls: 97900, tags: [ '1.8b' ], hasVision: true, added: '20240501' }, - 'neural-chat': { description: 'A fine-tuned model based on Mistral with good coverage of domain and language.', pulls: 97300, tags: [ '7b' ] }, + 'mistral-large': { description: 'Mistral Large 2 is Mistral\'s new flagship model that is significantly more capable in code generation, mathematics, and reasoning with 128k context window and support for dozens of languages.', pulls: 117900, tags: [ '123b' ], hasTools: true, added: '20241210' }, + 'command-r-plus': { description: 'Command R+ is a powerful, scalable large language model purpose-built to excel at real-world enterprise use cases.', pulls: 116200, contextWindow: 128000, tags: [ '104b' ], hasTools: true, added: '20240501' }, + 'stable-code': { description: 'Stable Code 3B is a coding model with instruct and code completion variants on par with models such as Code Llama 7B that are 2.5x larger.', pulls: 116000, tags: [ '3b' ] }, + 'openhermes': { description: 'OpenHermes 2.5 is a 7B model fine-tuned by Teknium on Mistral with fully open datasets.', pulls: 115900, tags: [] }, + 'tinydolphin': { description: 'An experimental 1.1B parameter model trained on the new Dolphin 2.8 dataset by Eric Hartford and based on TinyLlama.', pulls: 114600, tags: [ '1.1b' ] }, + 'glm4': { description: 'A strong multi-lingual general language model with competitive performance to Llama 3.', pulls: 114400, tags: [ '9b' ], added: '20241210' }, + 'wizardcoder': { description: 'State-of-the-art code generation model', pulls: 113700, tags: [ '33b' ] }, + 'qwen2-math': { description: 'Qwen2 Math is a series of specialized math language models built upon the Qwen2 LLMs, which significantly outperforms the mathematical capabilities of open-source models and even closed-source models (e.g., GPT4o).', pulls: 113100, tags: [ '1.5b', '7b', '72b' ], hasTools: true, added: '20241210' }, + 'bakllava': { description: 'BakLLaVA is a multimodal model consisting of the Mistral 7B base model augmented with the LLaVA architecture.', pulls: 104800, tags: [ '7b' ], hasVision: true }, + 'stablelm2': { description: 'Stable LM 2 is a state-of-the-art 1.6B and 12B parameter language model trained on multilingual data in English, Spanish, German, Italian, French, Portuguese, and Dutch.', pulls: 104500, tags: [ '1.6b', '12b' ] }, + 'deepseek-llm': { description: 'An advanced language model crafted with 2 trillion bilingual tokens.', pulls: 103100, tags: [ '7b', '67b' ] }, + 'reflection': { description: 'A high-performing model trained with a new technique called Reflection-tuning that teaches a LLM to detect mistakes in its reasoning and correct course.', pulls: 101800, tags: [ '70b' ], added: '20241210' }, + 'moondream': { description: 'moondream2 is a small vision language model designed to run efficiently on edge devices.', pulls: 98400, tags: [ '1.8b' ], hasVision: true, added: '20240501' }, + 'neural-chat': { description: 'A fine-tuned model based on Mistral with good coverage of domain and language.', pulls: 97700, tags: [ '7b' ] }, 'llama3-gradient': { description: 'This model extends LLama-3 8B\'s context length from 8k to over 1m tokens.', pulls: 95700, tags: [ '8b', '70b' ], added: '20240501' }, - 'wizard-math': { description: 'Model focused on math and logic problems', pulls: 95400, tags: [ '7b', '13b', '70b' ] }, - 'llama3-chatqa': { description: 'A model from NVIDIA based on Llama 3 that excels at conversational question answering (QA) and retrieval-augmented generation (RAG).', pulls: 87600, tags: [ '8b', '70b' ], added: '20240628' }, - 'deepseek-v2': { description: 'A strong, economical, and efficient Mixture-of-Experts language model.', pulls: 85800, tags: [ '16b', '236b' ], added: '20240628' }, - 'sqlcoder': { description: 'SQLCoder is a code completion model fined-tuned on StarCoder for SQL generation tasks', pulls: 85700, tags: [ '7b', '15b' ] }, + 'wizard-math': { description: 'Model focused on math and logic problems', pulls: 95500, tags: [ '7b', '13b', '70b' ] }, + 'llama3-chatqa': { description: 'A model from NVIDIA based on Llama 3 that excels at conversational question answering (QA) and retrieval-augmented generation (RAG).', pulls: 87700, tags: [ '8b', '70b' ], added: '20240628' }, + 'deepseek-v2': { description: 'A strong, economical, and efficient Mixture-of-Experts language model.', pulls: 87500, tags: [ '16b', '236b' ], added: '20240628' }, + 'sqlcoder': { description: 'SQLCoder is a code completion model fined-tuned on StarCoder for SQL generation tasks', pulls: 85800, tags: [ '7b', '15b' ] }, 'xwinlm': { description: 'Conversational model based on Llama 2 that performs competitively on various benchmarks.', pulls: 82900, tags: [ '7b', '13b' ] }, - 'minicpm-v': { description: 'A series of multimodal LLMs (MLLMs) designed for vision-language understanding.', pulls: 81400, tags: [ '8b' ], hasVision: true, added: '20241210' }, - 'nous-hermes': { description: 'General use models based on Llama and Llama 2 from Nous Research.', pulls: 80400, tags: [ '7b', '13b' ] }, - 'dolphincoder': { description: 'A 7B and 15B uncensored variant of the Dolphin model family that excels at coding, based on StarCoder2.', pulls: 80000, tags: [ '7b', '15b' ], added: '20240501' }, + 'minicpm-v': { description: 'A series of multimodal LLMs (MLLMs) designed for vision-language understanding.', pulls: 82000, tags: [ '8b' ], hasVision: true, added: '20241210' }, + 'nous-hermes': { description: 'General use models based on Llama and Llama 2 from Nous Research.', pulls: 80500, tags: [ '7b', '13b' ] }, + 'dolphincoder': { description: 'A 7B and 15B uncensored variant of the Dolphin model family that excels at coding, based on StarCoder2.', pulls: 80100, tags: [ '7b', '15b' ], added: '20240501' }, 'phind-codellama': { description: 'Code generation model based on Code Llama.', pulls: 79600, tags: [ '34b' ] }, 'yarn-llama2': { description: 'An extension of Llama 2 that supports a context of up to 128k tokens.', pulls: 77400, contextWindow: 128000, tags: [ '7b', '13b' ] }, - 'hermes3': { description: 'Hermes 3 is the latest version of the flagship Hermes series of LLMs by Nous Research', pulls: 75200, tags: [ '3b', '8b', '70b', '405b' ], hasTools: true, added: '20241210' }, - 'wizardlm': { description: 'General use model based on Llama 2.', pulls: 75000, tags: [] }, - 'solar': { description: 'A compact, yet powerful 10.7B large language model designed for single-turn conversation.', pulls: 74200, tags: [ '10.7b' ] }, - 'starling-lm': { description: 'Starling is a large language model trained by reinforcement learning from AI feedback focused on improving chatbot helpfulness.', pulls: 73700, tags: [ '7b' ] }, - 'yi-coder': { description: 'Yi-Coder is a series of open-source code language models that delivers state-of-the-art coding performance with fewer than 10 billion parameters.', pulls: 71700, tags: [ '1.5b', '9b' ], added: '20241210' }, - 'llava-phi3': { description: 'A new small LLaVA model fine-tuned from Phi 3 Mini.', pulls: 71100, tags: [ '3.8b' ], hasVision: true, added: '20240628' }, - 'internlm2': { description: 'InternLM2.5 is a 7B parameter model tailored for practical scenarios with outstanding reasoning capability.', pulls: 69100, tags: [ '1m', '1.8b', '7b', '20b' ], added: '20241210' }, - 'athene-v2': { description: 'Athene-V2 is a 72B parameter model which excels at code completion, mathematics, and log extraction tasks.', pulls: 68400, tags: [ '72b' ], hasTools: true, added: '20241210' }, - 'samantha-mistral': { description: 'A companion assistant trained in philosophy, psychology, and personal relationships. Based on Mistral.', pulls: 66500, tags: [ '7b' ] }, - 'falcon': { description: 'A large language model built by the Technology Innovation Institute (TII) for use in summarization, text generation, and chat bots.', pulls: 66500, tags: [ '7b', '40b', '180b' ] }, - 'mistral-small': { description: 'Mistral Small is a lightweight model designed for cost-effective use in tasks like translation and summarization.', pulls: 65100, tags: [ '22b' ], hasTools: true, added: '20241210' }, + 'hermes3': { description: 'Hermes 3 is the latest version of the flagship Hermes series of LLMs by Nous Research', pulls: 75500, tags: [ '3b', '8b', '70b', '405b' ], hasTools: true, added: '20241210' }, + 'wizardlm': { description: 'General use model based on Llama 2.', pulls: 75100, tags: [] }, + 'solar': { description: 'A compact, yet powerful 10.7B large language model designed for single-turn conversation.', pulls: 74500, tags: [ '10.7b' ] }, + 'starling-lm': { description: 'Starling is a large language model trained by reinforcement learning from AI feedback focused on improving chatbot helpfulness.', pulls: 74000, tags: [ '7b' ] }, + 'yi-coder': { description: 'Yi-Coder is a series of open-source code language models that delivers state-of-the-art coding performance with fewer than 10 billion parameters.', pulls: 71900, tags: [ '1.5b', '9b' ], added: '20241210' }, + 'llava-phi3': { description: 'A new small LLaVA model fine-tuned from Phi 3 Mini.', pulls: 71500, tags: [ '3.8b' ], hasVision: true, added: '20240628' }, + 'internlm2': { description: 'InternLM2.5 is a 7B parameter model tailored for practical scenarios with outstanding reasoning capability.', pulls: 69300, tags: [ '1m', '1.8b', '7b', '20b' ], added: '20241210' }, + 'mistral-small': { description: 'Mistral Small 3 sets a new benchmark in the โ€œsmallโ€ Large Language Models category below 70B.', pulls: 69100, tags: [ '22b', '24b' ], hasTools: true, added: '20250130' }, + 'athene-v2': { description: 'Athene-V2 is a 72B parameter model which excels at code completion, mathematics, and log extraction tasks.', pulls: 69100, tags: [ '72b' ], hasTools: true, added: '20241210' }, + 'samantha-mistral': { description: 'A companion assistant trained in philosophy, psychology, and personal relationships. Based on Mistral.', pulls: 66600, tags: [ '7b' ] }, + 'falcon': { description: 'A large language model built by the Technology Innovation Institute (TII) for use in summarization, text generation, and chat bots.', pulls: 66600, tags: [ '7b', '40b', '180b' ] }, 'orca2': { description: 'Orca 2 is built by Microsoft research, and are a fine-tuned version of Meta\'s Llama 2 models. The model is designed to excel particularly in reasoning.', pulls: 60800, tags: [ '7b', '13b' ] }, - 'nemotron': { description: '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.', pulls: 59900, tags: [ '70b' ], hasTools: true, added: '20241210' }, - 'dolphin-phi': { description: 'A 2.7B uncensored Dolphin model by Eric Hartford, based on the Phi language model by Microsoft Research.', pulls: 57900, tags: [ '2.7b' ] }, - 'stable-beluga': { description: 'Llama 2 based model fine tuned on an Orca-style dataset. Originally called Free Willy.', pulls: 57800, tags: [ '7b', '13b', '70b' ] }, - 'wizardlm-uncensored': { description: 'Uncensored version of Wizard LM model', pulls: 56700, tags: [ '13b' ] }, - 'nemotron-mini': { description: 'A commercial-friendly small language model by NVIDIA optimized for roleplay, RAG QA, and function calling.', pulls: 56700, tags: [ '4b' ], hasTools: true, added: '20241210' }, - 'llama3-groq-tool-use': { description: 'A series of models from Groq that represent a significant advancement in open-source AI capabilities for tool use/function calling.', pulls: 45500, tags: [ '8b', '70b' ], hasTools: true, added: '20241210' }, + 'nemotron': { description: '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.', pulls: 60000, tags: [ '70b' ], hasTools: true, added: '20241210' }, + 'dolphin-phi': { description: 'A 2.7B uncensored Dolphin model by Eric Hartford, based on the Phi language model by Microsoft Research.', pulls: 58200, tags: [ '2.7b' ] }, + 'stable-beluga': { description: 'Llama 2 based model fine tuned on an Orca-style dataset. Originally called Free Willy.', pulls: 57900, tags: [ '7b', '13b', '70b' ] }, + 'nemotron-mini': { description: 'A commercial-friendly small language model by NVIDIA optimized for roleplay, RAG QA, and function calling.', pulls: 57500, tags: [ '4b' ], hasTools: true, added: '20241210' }, + 'wizardlm-uncensored': { description: 'Uncensored version of Wizard LM model', pulls: 56800, tags: [ '13b' ] }, + 'llama3-groq-tool-use': { description: 'A series of models from Groq that represent a significant advancement in open-source AI capabilities for tool use/function calling.', pulls: 45700, tags: [ '8b', '70b' ], hasTools: true, added: '20241210' }, + 'granite3-dense': { description: 'The IBM Granite 2B and 8B models are designed to support tool-based use cases and support for retrieval augmented generation (RAG), streamlining code generation, translation and bug fixing.', pulls: 44600, tags: [ '2b', '8b' ], hasTools: true, added: '20241210' }, 'llama-pro': { description: 'An expansion of Llama 2 that specializes in integrating both general language understanding and domain-specific knowledge, particularly in programming and mathematics.', pulls: 44400, tags: [] }, + 'medllama2': { description: 'Fine-tuned Llama 2 model to answer medical questions based on an open source medical dataset.', pulls: 44400, tags: [ '7b' ] }, 'yarn-mistral': { description: 'An extension of Mistral to support context windows of 64K or 128K.', pulls: 44300, tags: [ '7b' ] }, - 'medllama2': { description: 'Fine-tuned Llama 2 model to answer medical questions based on an open source medical dataset.', pulls: 44300, tags: [ '7b' ] }, - 'granite3-dense': { description: 'The IBM Granite 2B and 8B models are designed to support tool-based use cases and support for retrieval augmented generation (RAG), streamlining code generation, translation and bug fixing.', pulls: 44200, tags: [ '2b', '8b' ], hasTools: true, added: '20241210' }, - 'meditron': { description: 'Open-source medical large language model adapted from Llama 2 to the medical domain.', pulls: 43500, tags: [ '7b', '70b' ] }, - 'granite3.1-dense': { description: 'The IBM Granite 2B and 8B models are text-only dense LLMs trained on over 12 trillion tokens of data, demonstrated significant improvements over their predecessors in performance and speed in IBMโ€™s initial testing.', pulls: 43100, tags: [ '2b', '8b' ], hasTools: true, added: '20250128' }, + 'granite3.1-dense': { description: 'The IBM Granite 2B and 8B models are text-only dense LLMs trained on over 12 trillion tokens of data, demonstrated significant improvements over their predecessors in performance and speed in IBMโ€™s initial testing.', pulls: 43700, tags: [ '2b', '8b' ], hasTools: true, added: '20250128' }, + 'meditron': { description: 'Open-source medical large language model adapted from Llama 2 to the medical domain.', pulls: 43700, tags: [ '7b', '70b' ] }, 'nexusraven': { description: 'Nexus Raven is a 13B instruction tuned model for function calling tasks.', pulls: 40600, tags: [ '13b' ] }, - 'dolphin3': { description: 'Dolphin 3.0 Llama 3.1 8B ๐Ÿฌ is the next generation of the Dolphin series of instruct-tuned models designed to be the ultimate general purpose local model, enabling coding, math, agentic, function calling, and general use cases.', pulls: 38100, tags: [ '8b' ], added: '20250128' }, - 'nous-hermes2-mixtral': { description: 'The Nous Hermes 2 model from Nous Research, now trained over Mixtral.', pulls: 37200, tags: [ '8x7b' ] }, + 'dolphin3': { description: 'Dolphin 3.0 Llama 3.1 8B ๐Ÿฌ is the next generation of the Dolphin series of instruct-tuned models designed to be the ultimate general purpose local model, enabling coding, math, agentic, function calling, and general use cases.', pulls: 38800, tags: [ '8b' ], added: '20250128' }, + 'nous-hermes2-mixtral': { description: 'The Nous Hermes 2 model from Nous Research, now trained over Mixtral.', pulls: 37300, tags: [ '8x7b' ] }, 'codeup': { description: 'Great code generation model based on Llama2.', pulls: 36000, tags: [ '13b' ] }, - 'smallthinker': { description: 'A new small reasoning model fine-tuned from the Qwen 2.5 3B Instruct model.', pulls: 35400, tags: [ '3b' ], added: '20250128' }, + 'smallthinker': { description: 'A new small reasoning model fine-tuned from the Qwen 2.5 3B Instruct model.', pulls: 35700, tags: [ '3b' ], added: '20250128' }, 'everythinglm': { description: 'Uncensored Llama2 based model with support for a 16K context window.', pulls: 35200, tags: [ '13b' ] }, - 'aya-expanse': { description: 'Cohere For AI\'s language models trained to perform well across 23 different languages.', pulls: 31600, tags: [ '8b', '32b' ], hasTools: true, added: '20241210' }, + 'aya-expanse': { description: 'Cohere For AI\'s language models trained to perform well across 23 different languages.', pulls: 31900, tags: [ '8b', '32b' ], hasTools: true, added: '20241210' }, 'magicoder': { description: '๐ŸŽฉ Magicoder is a family of 7B parameter models trained on 75K synthetic instruction data using OSS-Instruct, a novel approach to enlightening LLMs with open-source code snippets.', pulls: 31100, tags: [ '7b' ] }, - 'falcon2': { description: 'Falcon2 is an 11B parameters causal decoder-only model built by TII and trained over 5T tokens.', pulls: 30800, tags: [ '11b' ], added: '20240628' }, + 'falcon2': { description: 'Falcon2 is an 11B parameters causal decoder-only model built by TII and trained over 5T tokens.', pulls: 30900, tags: [ '11b' ], added: '20240628' }, 'stablelm-zephyr': { description: 'A lightweight chat model allowing accurate, and responsive output without requiring high-end hardware.', pulls: 30500, tags: [ '3b' ] }, 'codebooga': { description: 'A high-performing code instruct model created by merging two existing code models.', pulls: 30200, tags: [ '34b' ] }, 'granite3-moe': { description: 'The IBM Granite 1B and 3B models are the first mixture of experts (MoE) Granite models from IBM designed for low latency usage.', pulls: 29900, tags: [ '1b', '3b' ], hasTools: true, added: '20241210' },