Compare Models
-
Microsoft, NVIDIA
MT-NLG
OTHERMT-NLG (Megatron-Turing Natural Language Generation) uses the architecture of the transformer-based Megatron to generate coherent and contextually relevant text for a range of tasks, including completion prediction, reading comprehension, commonsense reasoning, natural language inferences, and word sense disambiguation. MT-NLG is the successor to Microsoft Turing NLG 17B and NVIDIA Megatron-LM 8.3B. The MT-NLG model is three times larger than GPT-3 (530B vs 175B). Following the original Megatron work, NVIDIA and Microsoft trained the model on over 4,000 GPUs. NVIDIA has announced an Early Access program for its managed API service to the MT-NLG model for organizations and researchers. -
Google
PaLM 2 chat-bison-001
$0.0021535PaLM 2 has just launched (May 2023) and is Google’s next-generation Large Language Model, built on Google’s Pathways AI architecture. PaLM 2 was trained on a massive dataset of text and code, and it can handle many different tasks and learn new ones quickly. It is seen as a direct competitor to OpenAI’s GPT-4 model. It excels at advanced reasoning tasks, including code and math, classification and question answering, translation and multilingual proficiency (100 languages), and natural language generation better than our previous state-of-the-art LLMs, including its predecessor PaLM.PaLM 2 is the underlying model driving the PaLM API that can be accessed through Google’s Generative AI Studio. PaLM 2 has four submodels with different sizes. Bison is the best value in terms of capability and chat-bison-001 has been fine-tuned for multi-turn conversation use cases. If you want to see PaLM 2 capabilities, the simplest way to use it is through Google Bard (PaLM 2 is the technology that powers Google Bard).Watch Paige Bailey introducing PaLM 2: view here
-
ChatGLM
PaLM 2 text-bison-001
$0.004PaLM 2 has just launched (May 2023) and is Google’s next-generation Large Language Model, built on Google’s Pathways AI architecture. PaLM 2 was trained on a massive dataset of text and code, and it can handle many different tasks and learn new ones quickly. It is seen as a direct competitor to OpenAI’s GPT-4 model. It excels at advanced reasoning tasks, including code and math, classification, question answering, translation and multilingual proficiency (100 languages), and natural language generation better than our previous state-of-the-art LLMs, including its predecessor PaLM.PaLM 2 is the underlying model driving the PaLM API that can be accessed through Google’s Generative AI Studio. PaLM 2 has four submodels with different sizes. Bison is the best value in terms of capability and cost, and text-bison-001 can be fine-tuned to follow natural language instructions and is suitable for various language tasks such as classification, sentiment analysis, entity extraction, extractive question answering, summarization, re-writing text in a different style, and concept ideation.If you want to see PaLM 2 capabilities, the simplest way to use it is through Google Bard (PaLM 2 is the technology that powers Google Bard).
Watch Paige Bailey introducing PaLM 2: view here
-
Google
PaLM 2 textembedding-gecko-001
$0.0004PaLM 2 has just launched (May 2023) and is Google’s next-generation Large Language Model, built on Google’s Pathways AI architecture. PaLM 2 was trained on a massive dataset of text and code, and it can handle many different tasks and learn new ones quickly. It is seen as a direct competitor to OpenAI’s GPT-4 model. It excels at advanced reasoning tasks, including code and math, classification and question answering, translation and multilingual proficiency (100 languages), and natural language generation better than our previous state-of-the-art LLMs, including its predecessor PaLM.PaLM 2 is the underlying model driving the PaLM API that can be accessed through Google’s Generative AI Studio. PaLM 2 has four submodels with different sizes: Unicorn (the largest), Bison, Otter, and Gecko (the smallest) and the different sizes of the submodels allow PaLM 2 to be more efficient and to perform different tasks. Gecko is the smallest and cheapest model for simple tasks and textembedding-gecko-001 returns model embeddings for text inputs.If you want to see PaLM 2 capabilities, the simplest way to use it is through Google Bard (PaLM 2 is the technology that powers Google Bard).Watch Paige Bailey introducing PaLM 2: view here
-
StableLM
StableLM-Base-Alpha -7B
FREEStability AI released a new open-source language model, StableLM. The Alpha version of the model is available in 3 billion and 7 billion parameters. StableLM is trained on a new experimental dataset built on The Pile, but three times larger with 1.5 trillion tokens of content. The richness of this dataset gives StableLM surprisingly high performance in conversational and coding tasks, despite its small size. The models are now available on GitHub and on Hugging Face, and developers can freely inspect, use, and adapt our StableLM base models for commercial or research purposes subject to the terms of the CC BY-SA-4.0 license.
-
OpenAI
text-embedding-ada-002
$0.0001An embedding API model, such as Ada, is a powerful tool that converts words into numerical representations, enabling computers to understand and process natural language more effectively. This process is crucial for developing machine learning algorithms and artificial intelligence systems that can interact with humans, analyze text, or make predictions based on text. OpenAI’s text embeddings is built for advanced search, clustering, topic modeling, and classification functionality.Access is available through a request to OpenAI’s API. -
LMSYS Org
Vicuna-13B
FREEVicuna-13B is an open-source chatbot developed by a team of researchers from UC Berkeley, CMU, Stanford, MBZUAI, and UC San Diego. The chatbot was trained by fine-tuning LLaMA on user-shared conversations collected from ShareGPT. There is a 13B and 7B parameter models that are available on Hugging Face.
Vicuna-13B achieves more than 90% quality of OpenAI ChatGPT and Google Bard while outperforming other models like LLaMA and Stanford Alpaca in more than 90% of cases. The code and weights and an online demo are publicly available for non-commercial use. Here is a link to learn more about how it compares to other models – https://lmsys.org/blog/2023-03-30-vicuna/.
To use this model, you need to install LLaMA weights first and convert them into Hugging Face weights, and the cost of training Vicuna-13B is around $300.
-
Yandex
YaLM
FREEYaLM 100B is a GPT-like neural network for generating and processing text. It can be used freely by developers and researchers from all over the world. It took 65 days to train the model on a cluster of 800 A100 graphics cards and 1.7 TB of online texts, books, and countless other sources in both English and Russian. Researchers and developers can use the corporate-size solution to solve the most complex problems associated with natural language processing.Training details and best practices on acceleration and stabilizations can be found on Medium (English) and Habr (Russian) articles. The model is published under the Apache 2.0 license that permits both research and commercial use.
