273 Ventures

273 Ventures, is a company that develops a Legal Data Operating System (OS) to organize and connect data from various sources within legal organizations, which includes document sockets and other sources.&nbsp;

Foundation Models

In a Gen AI First, 273 Ventures Introduces KL3M, a Built-From-Scratch Legal LLM

273 Ventures, is a company that develops a Legal Data Operating System (OS) to organize and connect data from various sources within legal organizations, which includes document sockets and other sources.&nbsp;The company’s flagship model, KL3M, is a family of large language models (LLMs) specifically trained for applications in the legal sector and broader enterprise use, capable of tasks like answering regulatory questions, drafting contract provisions, extracting information, and executing direct preference optimization. Built from a custom dataset with full legal clearance, these models are designed to meet diverse requirements across different generations, sizes, and specialized purposes. According to the company, the initial models in the KL3M series aim to deliver highly efficient foundation models that support downstream tasks, including optimization, alignment, and fine-tuning for specific organizational applications.&nbsp;Its variations&nbsp; (launched in&nbsp;<a href="https://sp-edge.com/updates/26834">February 2024)</a> are KL3m-170m and KL3m-1.7b, which claim to reduce litigation complications and ethical concerns regarding data origins.&nbsp;<h3>Key Customers and Partnerships&nbsp;</h3>The models were available for Kelvin Legal Data OS customers as of February 2024, with general availability planned in the future.

AI21 is an Israel-based AI research lab developing large language models (LLMs). The firm also builds GenAI applications based on its LLMs, marketing them to businesses and consumers.&nbsp;AI21 Labs has developed several language models since its inception, beginning with Jurassic-1 (August 2021) and progressing through Jurassic-2 (March 2023) and&nbsp;<a href="https://sp-edge.com/updates/27995">Jamba</a> (March 2024). The company offers&nbsp;<a href="https://sp-edge.com/updates/32879">Jamba 1.5 Mini and Jamba 1.5 Large,</a> launched in August 2024, marking a significant advancement in long-context language processing. These models support a 256,000-token context window and are optimized for speed, providing up to 2.5x faster inference on extended contexts. Jamba 1.5 Large is a Mixture-of-Experts (MoE) model with 398 billion total parameters and 94 billion active parameters, while Jamba 1.5 Mini is an enhanced version of Jamba-instruct, emphasizing efficiency in handling complex language tasks. The company claims that its foundation models are 30% more efficient than those of other providers and it does not charge additional fees for the use of fine-tuned models.&nbsp;Additionally, AI21 offers AI21 Studio, a platform for custom model training and API integration, and has developed the MRKL framework to enhance LLMs with external knowledge sources and reasoning capabilities.<h3>Key customers and partnerships</h3>As of June 2023, AI21’s customers include Latitude, Easyway, Verb, Tweet Hunter, and Harambee. In May 2024,&nbsp;<a href="https://sp-edge.com/updates/29800">the company partnered with Aimpoint Digital</a>, provider of analytics, and Weights and Biases, an integrated MLOps platform, to leverage AI21's GenAI models which are tuned for RAG for Aimpoint Digital's GenAI solutions. Further,&nbsp;<a href="https://sp-edge.com/updates/29663">AI21 partnered with Microsoft&nbsp;</a>to incorporate AI21's latest model, Jamba-Instruct, into Microsoft's Azure AI as part of its Model-as-a-Service (MaaS) offering.&nbsp;

AI21 Labs

UK-based Stability AI builds AI applications and offers a range of&nbsp;foundation models.&nbsp;It has created several text-to-image models, including Stable Diffusion XL (SDXL 1.0), which utilizes a dual-network architecture with a 3.5-billion-parameter base model and a 6.6-billion-parameter refiner for high-quality image generation, and&nbsp;<a href="https://sp-edge.com/updates/30420">Stable Diffusion 3 Medium</a>, an open-source model with 2 billion parameters designed for fine-tuning with small datasets.Additionally,&nbsp;<a href="https://sp-edge.com/updates/32260">Stable Fast 3D</a> generates 3D images from single 2D inputs. The company also offers Stable Diffusion 3 Large and Stable Diffusion 3 Large Turbo, along with Stable Diffusion 3.5 Medium (2.5 billion parameters), Stable Diffusion 3.5 Large (8 billion parameters), and Stable Diffusion 3.5 Large Turbo, a distilled version of the latter. In November 2023,&nbsp;<a href="https://sp-edge.com/updates/23352">the company launched Sky Replacer, which allows users to modify the sky in photos, and Stable 3D</a>, a tool for creating textured 3D objects from images or text prompts.Stability AI has also released several text-to-text models, including the Stable Beluga series, StableLM series, which includes open-source language models ranging from three billion to 65 billion parameters for text and code generation, featuring models like Stable LM 2 12B, and Stable LM 2 1.6B, a smaller multilingual model.&nbsp;<a href="https://sp-edge.com/updates/36984">Stable LM 1.6 billion</a> is an Arabic language model available in base and chat versions, which has been fine-tuned on over 100 billion Arabic text tokens and uses the Arcade100k tokenizer. The Japanese Stable LM, considered sophisticated to other publicly available Japanese language models, is optimized for Japanese text generation, while StableCode (three billion parameters) provides autocomplete suggestions for programming tasks.For audio generation, Stability AI has developed Stable Audio, which uses a latent diffusion model architecture to generate audio based on text prompts, along with its successor, Stable Audio 2.0, optimized for short audio samples and sound effects.In video generation, Stable Video Diffusion produces short clips from images, while&nbsp;<a href="https://sp-edge.com/updates/27695">Stable Video 3D (SV3D)</a> creates multi-view 3D models from single images or text prompts, and&nbsp;<a href="https://sp-edge.com/updates/31945">Stable Video 4D</a> enables users to upload videos to obtain dynamic outputs from various perspectives. These are also adaptable to various video applications and can be fine-tuned for multi-view synthesis using multi-view datasets. In May 2024, Stability AI introduced Stable Artisan, a tool designed for media generation within Discord environments. Key customers and partnerships&nbsp;In December 2022, Stability AI partnered with Amazon, with AWS designated as the preferred cloud provider to facilitate the development and expansion of Stability AI's AI models.&nbsp;<a href="https://sp-edge.com/updates/27403">Stability AI partnered with Tripo AI</a> in March 2024 to release TripoSR for 3D object reconstruction to cater to the demands of entertainment, gaming, industrial design, and architecture professionals, with responsive outputs for visualizing detailed 3D objects. In April 2024,&nbsp;the company&nbsp;<a href="https://sp-edge.com/updates/28612">partnered with Fireworks AI</a>, a provider of API platforms, to deliver Stable Diffusion 3 and Stable Diffusion 3 Turbo for enterprises.&nbsp;In December 2024,&nbsp;<a href="https://sp-edge.com/updates/37283">the company partnered with COLOPL&nbsp;</a>to integrate AI technology into game development, specifically focusing on using Stable Diffusion for image generation.

Stability AI

Cohere is a Canadian startup that develops and deploys large language models (LLMs) for various applications across enterprise and business use cases.&nbsp;Cohere's models include: (1) Command R (March 2024), to handle large-scale production workloads for global languages in businesses; (2) Command R+ (April 2024), an enhanced version of Command R and offers enhanced retrieval augmented generation (RAG) with citation, supports 10 business languages, and automates workflows; (3)&nbsp;<a href="https://sp-edge.com/updates/29350">Command R Fine-Tuning</a> (May 2024), for enhanced performance for business application workload; (4)&nbsp;<a href="https://sp-edge.com/updates/37280">Command R7B</a>, its smallest and fastest language model in the R series, which features a context length of 128K and supports 23 languages; (5)&nbsp;<a href="https://sp-edge.com/updates/28463">Rerank 3 (April 2024)</a>, a model to enhance enterprise search and retrieval augmented generation (RAG) systems and has a 4k context length for better search results in longer documents, the ability to examine diverse and semi-structured data like emails, JSON files, invoices, and programming codes, multilingual capabilities across 100+ languages, improved latency, and lower total cost of ownership; (6)&nbsp;<a href="https://sp-edge.com/updates/36743">Rerank 3.5 (November 2024)</a>, an enterprise search model that can process queries in 100+ languages; (7)&nbsp;<a href="https://sp-edge.com/updates/29761">Aya 23 (May 2024)</a>, a multilingual language model that can handle 23 languages&nbsp; (8) Embed 3 (October 2024), a multimodal AI search model that can generate embeddings from text and images<h3>Key customers and partnerships&nbsp;</h3>As of December 2024, key customers of the company include Oracle, Johnson Lambert, Atomicwork, Borderless AI, BlueDot, and HyperWrite.Cohere has established a robust partner ecosystem. In 2024, Cohere has partnered with 1)&nbsp;<a href="https://sp-edge.com/updates/27307">Accenture</a> (March) to integrate its Command model and retrieval augmented generation (RAG) capabilities to offer GenAI solutions for businesses, 2)&nbsp;<a href="https://sp-edge.com/updates/28932">Amazon&nbsp;</a>(April) to incorporate Command R+ and Command R to Amazon's cloud platform, Bedrock, 3)&nbsp;<a href="https://sp-edge.com/updates/30492">Borderless AI</a> (June), a provider of AI-powered HR solutions, to create tailored AI agents for HR tasks, 4)&nbsp;<a href="https://sp-edge.com/updates/31052">i-5O</a> (July), a provider of AI solutions for manufacturing, as part of Cohere's Startup Program, 5)&nbsp;<a href="https://sp-edge.com/updates/31590">Fujitsu</a> (July) to develop a Japanese LLM, 6)&nbsp;<a href="https://sp-edge.com/updates/33458">Nomura Research Institute (NRI)</a> (September), a Japanese consulting firm, to develop the NRI Financial AI Platform, and&nbsp;7)&nbsp;<a href="https://sp-edge.com/updates/37310">Palantir</a> (December) to deploy its AI models through Palantir's Foundry platform.

Cohere

Hippocratic AI develops a safety-focused large language model (LLM) specifically designed for healthcare applications, with an emphasis on non-diagnostic, patient-facing tasks.&nbsp;Hippocratic AI aims to differentiate itself from traditional LLMs by prioritizing pre-training with trusted, evidence-based healthcare content rather than the broader Internet.&nbsp;Claimed to be the first safety-focused LLM for healthcare, the model undergoes a unique Reinforcement Learning with Human Feedback process, known as RLHF-HP, which involves working with healthcare professionals to validate its readiness for use.&nbsp;The company offers a staffing marketplace, where healthcare providers can “hire” AI agents at USD 9 per hour to perform various low-risk tasks such as chronic care management, post-discharge follow-up for specific conditions like congestive heart failure and kidney disease, wellness surveys, health risk assessments, and pre-operative outreach. In addition, the model aims to incorporate a compassionate “bedside manner” in its interactions with patients, recognizing the impact of empathy on emotional wellbeing and health outcomes.<h3>Key customers and partnerships&nbsp;</h3>In August 2023, the company launched the Hippocratic AI Founding Partner Program, an initiative through which the company collaborates with leading healthcare organizations to become founding partners and actively participate in the development and validation of Hippocratic AI's LLM and technology. The program includes establishing use cases, ensuring clinical and patient safety, and contributing to the advancement of healthcare-specific AI applications. In February 2024, these companies including Memorial Hermann Health System, University of Vermont Health Network, Fraser Health, and Side Health, began&nbsp;<a href="https://sp-edge.com/updates/27145">beta</a> testing their LLM for non-diagnostic applications in healthcare.<a href="https://sp-edge.com/updates/27696">Hippocratic AI partnered with NVIDIA</a> in March 2024 to develop AI healthcare agents that have low-latency conversational interactions with patients. In September 2024, it partnered with Adtalem Global Education to develop a suite of educational materials for training clinicians on the use of AI in healthcare.

Hippocratic AI

Founded in 2023, GenHealth.ai is a startup focused on advancing healthcare solutions through the use of GenAI.&nbsp;Its proprietary approach involves training a foundation transformer model, known as DOOG-E, on extensive healthcare data, including 50+ million of coded patient claims and clinical information. The model has 70 million parameters and one billion tokens. Unlike traditional large language models, DOOG-E is a large medical model (LMM) that uses medical events such as ICD, CPT, RxNorm, and LOINC codes as its "language" to predict future sequences of health events based on historical data. In September 2024, GenHealth.ai launched the Large Medical Model (LMM), which, according to the company, integrates advanced technologies, including a healthcare-specific vocabulary tailored for medical applications, neural network transformers that power Large Language Models, and the analysis of data from trillions of healthcare events involving 140 million patients. Reportedly, it is trained on medical data and is less prone to bias and hallucination compared to its GenAI large language models (LLM), rendering it more dependable for healthcare applications.

GenHealth

Foundation models that can handle a range of natural language processing (NLP) use cases. Typical tasks include generating summaries, translating, answering questions, and classifying text. These models typically use text inputs and generate text outputs.

Fine-tuned versions of large language models, which are focused on specific tasks and applications (e.g. code generation, medical event prediction, financial NLP tasks, etc). These are usually developed by training LLMs using custom data and also serve as foundation models that can be developed even further. These models typically use text inputs and can generate text and/or code outputs.

Foundation models that translate visual data based on features and contextual information identified during training. This enables models to interpret images and video, applying those interpretations to predictive or decision-making tasks. Typical tasks include image recognition and classification into categories. These models typically use image and/or video inputs and generate text outputs.

Foundation models that are trained to understand and work with two or more forms of input data (e.g., text and images) to solve problems and complete tasks using both natural language processing and computer vision. These models can also generate outputs across textual and visual modalities. Typical tasks include visual question answering, image captioning, and generating images based on text descriptions. These models typically use text and/or visual inputs and generate text and/or visual outputs.

Foundation models that use big data to suggest or recommend additional products to consumers. These models are designed to predict user-item interaction probabilities and provide recommendations to users. The models are trained to make these inferences based on historical data on customer interactions and non-interactions. These models typically use user features, item features and user interactions as inputs, and generate numeric values like ratings predictions and matching probabilities as outputs.

Foundation models that are capable of automatic speech recognition, transcription in multiple languages, and translation from these languages to English. These are sometimes referred to as audio models. These models typically use speech/audio as inputs and generate text outputs.

Foundation models that transform user text prompts into video clips.

Foundation models that generate music from text descriptions.

All companies within the space are building large AI models

Companies in this space leverage NLP to train LLMs

The hub exclusively covers companies that have built foundation models

All companies within the space can leverage their foundation models for generative AI applications

The technology enables the creation of personalized content

273 Ventures

Key Customers and Partnerships

About EDGE

Contact Us

Resources

Area of focus

Technology Research Made Effortless