In the ever-evolving world of artificial intelligence (AI), open-source models have emerged as a game-changer. These models provide an opportunity for developers, researchers, and businesses to harness the power of AI, driving innovation and reducing the barriers to entry. The best part is that many of these models can be easily run locally on your hardware, allowing you to utilize AI in your projects without relying on cloud services.
Open-source models come in various flavors, with different sizes and complexities, making them suitable for a wide range of hardware configurations. Depending on your hardware capabilities, you can choose to run small, lightweight models or opt for more powerful, resource-intensive models. This flexibility ensures that everyone, from individual developers to large enterprises, can leverage the benefits of AI according to their unique needs and constraints.
For instance, if you have limited hardware resources, you can still take advantage of smaller, lightweight models that require minimal computational power. These models, while simpler in structure, can still deliver impressive results for many applications, such as text generation, translation, and basic conversational AI.
On the other hand, if you have access to high-performance hardware, you can unlock the full potential of more complex models. These models can offer advanced features, superior performance, and greater accuracy in handling intricate tasks. High-performance hardware, such as powerful CPUs, GPUs, or TPUs, can handle the increased computational demands of these models, ensuring smooth and efficient operation.
To harness the power of open-source models locally, consider utilizing solutions like Ollama or LM Studio (there are numerous tutos on the net - how to install and run). These platforms simplify the installation and management of open-source models, allowing you to run them seamlessly on your local machine. With user-friendly interfaces and compatibility with various models, Ollama and LM Studio streamline the process of integrating open-source models into your projects.
- Gemma: A family of lightweight, state-of-the-art open models built by Google DeepMind.
- Llama 2: A collection of foundation language models ranging from 7B to 70B parameters.
- Mistral: The 7B model released by Mistral AI, updated to version 0.2.
- Mixtral: A high-quality Mixture of Experts (MoE) model with open weights by Mistral AI.
- LLaVA: A novel end-to-end trained large multimodal model that combines a vision encoder and Vicuna for general-purpose visual and language understanding.
- Neural-Chat: A fine-tuned model based on Mistral with good coverage of domain and language.
- CodeLLAMA: A large language model that can use text prompts to generate and discuss code.
- Dolphin-Mixtral: An uncensored, fine-tuned model based on the Mixtral mixture of experts model that excels at coding tasks.
- Qwen 1.5: A series of large language models by Alibaba Cloud spanning from 0.5B to 72B parameters.
- Llama2-uncensored: An uncensored Llama 2 model by George Sung and Jarrad Hope.
- Mistral-OpenOrca: Mistral OpenOrca is a 7 billion parameter model, fine-tuned on top of the Mistral 7B model using the OpenOrca dataset.
- DeepSeek Coder: A capable coding model trained on two trillion code and natural language tokens.
- Nous-Hermes2: The powerful family of models by Nous Research that excels at scientific discussion and coding tasks.
- Phi-2: A 2.7B language model by Microsoft Research that demonstrates outstanding reasoning and language understanding capabilities.
- Orca-mini: A general-purpose model ranging from 3 billion parameters to 70 billion, suitable for entry-level hardware.
- Dolphin-Mistral: The uncensored Dolphin model based on Mistral that excels at coding tasks.
- Wizard-Vicuna-uncensored: Wizard Vicuna Uncensored is a 7B, 13B, and 30B parameter model based on Llama 2 uncensored.
- Vicuna: General use chat model based on Llama and Llama 2 with 2K to 16K context sizes.
- TinyDolphin: An experimental 1.1B parameter model trained on the new Dolphin 2.8 dataset.
- Nomic-embed-text: A high-performing open embedding model with a large token context window.
- Llama2-chinese: Llama 2 based model fine-tuned to improve Chinese dialogue ability.
- OpenHermes: OpenHermes 2.5 is a 7B model fine-tuned by Teknium on Mistral with fully open datasets.
- Zephyr: Zephyr beta is a fine-tuned 7B version of Mistral that was trained on on a mix of publicly available, synthetic datasets.
- TinyLlama: The TinyLlama project is an open endeavor to train a compact 1.1B Llama model on 3 trillion tokens.
- OpenChat: A family of open-source models trained on a wide variety of data, surpassing ChatGPT on various benchmarks.
- WizardCoder: State-of-the-art code generation model.
- Starcoder: StarCoder is a code generation model trained on 80+ programming languages.
- Phind-Codellama: Code generation model based on Code Llama.
- Starcoder2: The next generation of transparently trained open code LLMs that comes in three sizes: 3B, 7B and 15B parameters.
- Yi: A high-performing, bilingual language model.
- Orca2: Orca 2 is built by Microsoft research, and are a fine-tuned version of Meta's Llama 2 models.
- Falcon: A large language model built by the Technology Innovation Institute (TII) for use in summarization, text generation, and chat bots.
- Wizard-Math: Model focused on math and logic problems.
- Dolphin-Phi: 2.7B uncensored Dolphin model by Eric Hartford, based on the Phi language model by Microsoft Research.
- Starling-LM: Starling is a large language model trained by reinforcement learning from AI feedback focused on improving chatbot helpfulness.
- Nous-Hermes: General use models based on Llama and Llama 2 from Nous Research.
- Stable-Code: 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.
- MedLLAMA2: Fine-tuned Llama 2 model to answer medical questions based on an open source medical dataset.
- **BakLLAVA**: BakLLaVA is a multimodal model consisting of the Mistral 7B base model augmented with the LLaVA architecture.
- CodeUp: Great code generation model based on Llama2.
- WizardLM-uncensored: Uncensored version of Wizard LM model.
- Solar: A compact, yet powerful 10.7B large language model designed for single-turn conversation.
- EverythingLM: Uncensored Llama2 based model with support for a 16K context window.
- SQLCoder: SQLCoder is a code completion model fined-tuned on StarCoder for SQL generation tasks.
- DolphinCoder: An uncensored variant of the Dolphin model family that excels at coding, based on StarCoder2.
- Nous-Hermes2-Mixtral: The Nous Hermes 2 model from Nous Research, now trained over Mixtral.
- StableBeluga: Llama 2 based model fine tuned on an Orca-style dataset.
- Yarn-Mistral: An extension of Mistral to support context windows of 64K or 128K.
- StableLM-Zephyr: A lightweight chat model allowing accurate, and responsive output without requiring high-end hardware.
- Magicoder: 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.
- Yarn-Llama2: An extension of Llama 2 that supports a context of up to 128k tokens.
- Llama-Pro: An expansion of Llama 2 that specializes in integrating both general language understanding and domain-specific knowledge, particularly in programming and mathematics.
- DeepSeek-LLM: An advanced language model crafted with 2 trillion bilingual tokens.
- Wizard-Vicuna: Wizard Vicuna is a 13B parameter model based on Llama 2 trained by MelodysDreamj.
- CodeBooga: A high-performing code instruct model created by merging two existing code models.
- MistralLite: MistralLite is a fine-tuned model based on Mistral with enhanced capabilities of processing long contexts.
- All-MinILM: Embedding models on very large sentence level datasets.
- Nexus Raven: Nexus Raven is a 13B instruction tuned model for function calling tasks.
- Open-Orca-Platypus2: Merge of the Open Orca OpenChat model and the Garage-bAInd Platypus 2 model. Designed for chat and code generation.
- Goliath: A language model created by combining two fine-tuned Llama 2 70B models into one.
- Notux: A top-performing mixture of experts model, fine-tuned with high-quality data.
- MegaDolphin: MegaDolphin-2.2-120b is a transformation of Dolphin-2.2-70b created by interleaving the model with itself.
- Alfred: A robust conversational model designed for both chat and instruct use cases.
- XwinLM: Conversational model based on Llama 2 that performs competitively on various benchmarks.
- WizardLM: General use 70 billion parameter model based on Llama 2.
- DuckDB-NSQL: 7B parameter text-to-SQL model made by MotherDuck and Numbers Station.
- Notus: 7B chat model fine-tuned with high-quality data and based on Zephyr.
- UAE-Large: Large language model with 2 Pulls and 3 Tags.
These open-source models offer a wide range of capabilities, from general-purpose chatbots and text generation to domain-specific tasks such as code generation and text-to-SQL conversion. By exploring and utilizing these open-source models, developers and researchers can accelerate innovation and drive the AI ecosystem forward.