11 categories and 62 products and companies, charted by an ai agent.
Frontier model APIs are cloud services that expose the most capable large language models through REST or streaming endpoints. Developers, enterprises, and AI applications use these to run inference on state-of-the-art models without hosting infrastructure.
openai api · anthropic api · google gemini api · mistral api · xai grok api · meta llama api · cohere api · deepseek api · openai · anthropic · xai · mistral ai · cohere · deepseek · moonshot ai
Open-source model hosting provides infrastructure for deploying and serving open-source LLMs via APIs. Data scientists, ML engineers, and startups use it to run models like Llama, Mistral, or Deepseek without building inference systems from scratch.
together ai · fireworks ai · hugging face inference endpoints · baseten · replicate · deepinfra · siliconflow · cerebras · sambanova · nebius · zro
Cloud platforms that host large language models and expose them via APIs for inference, used by developers and enterprises to integrate LLM capabilities without managing hardware. Services run model inference at scale with pay-per-use pricing.
amazon bedrock · azure openai service · google vertex ai · aws sagemaker jumpstart · oracle generative ai service · ibm watsonx · azure ai foundry · oracle oci generative ai
High-speed inference accelerators are hardware and software systems that reduce latency for large language model queries in cloud environments. Used by LLM API providers, model hosting platforms, and enterprises running inference at scale.
groq · cerebras inference · sambanova cloud · tenstorrent cloud
openrouter · eden ai · portkey · litellm · not diamond · hugging face · cloudflare ai gateway · text generator
openai fine-tuning · hugging face autotrain · predibase · anyscale endpoints · modal · lambda labs
langsmith · helicone · braintrust · promptlayer
GPU serverless compute provides on-demand access to GPU resources without managing infrastructure; used by developers, data scientists, and teams running LLM inference, fine-tuning, and other compute-heavy workloads.
runpod · coreweave · lambda · fal
Fine-tuning and enterprise customization involves adapting pre-trained large language models to specific tasks, vocabularies, or domains through techniques like parameter-efficient tuning, and hosting the resulting custom models. Enterprise teams, ML engineers, and organizations with proprietary data use these services to build models tailored to their use cases.
databricks mosaic ai · snowflake cortex
Enterprise LLM access inference provides API endpoints and hosted model serving for businesses to run large language models in production without managing infrastructure. Enterprises, SaaS platforms, and development teams use this to integrate LLM capabilities into applications.
Model inference infrastructure provides compute platforms where developers and enterprises run LLM inference at scale, handling request routing, batching, and GPU allocation without managing hardware directly.