Model Platforms & APIs
51 tools in the n3os directory.
E2X
by E2X AI
E2X is one prepaid API for image, video and audio models from OpenAI, Google and ByteDance, priced per request with no subscription or minimum.
Hirundo
Hirundo removes the effect of particular training data, such as bad labels or copyrighted material, from an already-trained model without full retraining.
TensorOpera
TensorOpera, formerly FedML, is a platform for training and deploying large models, with federated learning across separate data holders as its speciality.
LAION
LAION is a non-profit that publishes open machine learning resources, most significantly LAION-5B, a dataset of billions of image-text pairs.
Multiverse Computing
Multiverse Computing builds quantum and quantum-inspired software for finance, energy and life sciences, including model compression for ordinary hardware.
Vectara
Vectara provides retrieval-augmented generation as a service, with an evaluation layer that scores how well an answer is supported by its sources.
Reka AI
Reka AI is a foundation model lab building the Reka family: Core as the largest, Flash tuned for speed and cost, and Edge sized for on-device use.
Google AI Studio
by Google
Google AI Studio is a free web environment for prototyping with the Gemini model family: running prompts, generating media and getting an API key.
Portkey
Portkey is a gateway and control plane for language model traffic: one OpenAI-compatible endpoint across many models, with fallbacks, retries and caching.
Eden AI
Eden AI is a single API connected to a wide range of AI engines, covering generative models alongside OCR, translation, speech to text and moderation.
Stepfun
Stepfun builds the Step family of multimodal foundation models, including Step-Audio for voice. It was founded by Jiang Daxin and is backed by Tencent.
Actualyze AI
Actualyze AI is a single OpenAI-compatible layer between an organisation's applications and the model providers that governs and secures every request.
Sapiom
Sapiom is an inference router exposed through an OpenAI-compatible API, metering every call and routing it across efficient, mid-tier and frontier models.
Hydra Host
Hydra Host rents dedicated, non-virtualised NVIDIA GPU servers for AI, HPC and data workloads, across a network of more than forty data centres.
Bright Data
Bright Data provides infrastructure for collecting public web data: a proxy network, scraper APIs that return structured results, and pre-built datasets.
Invisible Technologies
Invisible Technologies supplies training data and AI operations, pairing a large expert workforce with software to produce data that frontier labs use.
MongoDB Atlas Vector Search
Atlas Vector Search adds semantic vector search to MongoDB Atlas. Embeddings are stored as fields on documents alongside the data they relate to.
Palantir AIP
Palantir AIP lets organisations use large language models on their private networks and their own operational data, within existing access controls.
Deccan AI
Deccan AI supplies expert-built training data for large language models and agents: fine-tuning datasets, RLHF preference data and model evaluations.
Voltage Park
Voltage Park rents NVIDIA H100 and Blackwell clusters for AI training and inference, either on demand by the hour or on longer dedicated-reserve contracts.
Datacurve
Datacurve supplies coding and engineering training data to frontier AI labs, sourced through Shipd, a bounty platform that pays software engineers.
Nebius
Nebius operates NVIDIA GPU clusters for training and inference, with managed Slurm and Kubernetes and an inference API on top. It is listed on Nasdaq.
Sakana AI
Sakana AI is a research lab founded by David Ha, formerly of Google Brain, and Llion Jones, a co-author of the original Transformer paper.
Imbue
Imbue is an AI research company working on agents that can reason and act safely on goals larger than a single prompt.
Auriko
Auriko is an LLM gateway: one OpenAI-compatible endpoint in front of more than a dozen model providers, routing requests by price, latency and reliability.
Tenstorrent
Tenstorrent builds AI training and inference accelerators on an open RISC-V architecture. It is led by Jim Keller and licenses its designs as well.
Etched
Etched builds Sohu, an application-specific integrated circuit designed to do one thing: transformer inference. It targets inference, not training.
Inception Labs
Inception Labs is a Stanford spinout building Mercury, a family of diffusion-based language models that refine the whole output in parallel.
Fireworks AI
Fireworks AI is a training and inference platform for open models, offering serverless and dedicated serving, fine-tuning and per-request model routing.
Cohere
Cohere builds and hosts its own models for generation, embeddings and reranking. They run in Cohere's cloud, your cloud account or your own hardware.
Cerebras
Cerebras builds AI accelerators on a single wafer rather than many separate chips, and sells access as a hosted inference service running open models.
RunPod
RunPod rents GPU compute in two shapes: pods, which are on-demand machines, and serverless endpoints, which start on a request and stop afterwards.
AI21 Labs
AI21 builds foundation models for business use, with a gateway that routes traffic to cut token cost and services that build a model on a customer's data.
Lambda
Lambda rents NVIDIA GPU compute: on-demand instances by the hour, one-click multi-node clusters, and reserved capacity for long runs.
Together AI
Together AI hosts a wide library of open models behind serverless and dedicated endpoints, with batch inference, fine-tuning and GPU clusters.
Groq
Groq runs open models on processors it designed for inference rather than training, and sells access as an API with an OpenAI-compatible interface.
SambaNova
SambaNova runs open models on reconfigurable dataflow hardware of its own design, sold as a hosted cloud service, a software stack and physical racks.
xAI Grok API
The Grok API provides developer access to xAI's models for text, reasoning, image and voice, with real-time access to posts on X as a data source.
Vertex AI
Vertex AI is Google Cloud's machine learning platform, giving access to Google's models and others with training, tuning, deployment and governance.
OpenAI
OpenAI is the research company behind ChatGPT and the GPT family of models. It provides a consumer assistant, a developer platform and media generation.
OpenAI API Platform
This is OpenAI's developer platform, not the consumer assistant. It provides API access to its models for text, images, audio and embeddings.
Fal.ai
Fal.ai runs generative media models for image, video and audio behind a fast API, concentrating on inference speed and one interface to many models.
Replicate
Replicate runs open-source models through an API. You call a hosted version of a published model without provisioning hardware, paying for compute used.
Llama
by Meta AI
Llama is Meta's family of open-weight language models. The weights can be downloaded and run on your own hardware or through many providers.
Meta AI
Meta AI is Meta's consumer assistant. It answers questions, generates images and helps with everyday tasks, and it is built into WhatsApp and Instagram.
Anthropic
Anthropic is an AI research company. It builds the Claude family of models, reachable through a chat interface, an API and a command-line coding tool.
Mistral
Mistral is a European AI company producing open-weight models and a commercial platform, with an API, a chat assistant and self-hosted deployment.
Hugging Face
Hugging Face is the main hub for open machine learning. It hosts models, datasets and demonstration applications, and offers hosted inference and training.
LM Studio
LM Studio is a desktop application for running language models on your own computer. It provides a chat interface and downloads models for you.
Ollama
Ollama runs open language models locally through a single command. It handles downloading, setup and serving, and exposes a local API for other software.
DeepSeek Platform
DeepSeek provides access to its language and coding models through an API and a chat interface. Several of its models have been released with open weights.