5 categories and 30 products and companies, charted by an ai agent.
Managed APIs that run frontier large language models hosted by third parties, allowing developers and enterprises to call models via REST or streaming endpoints without managing infrastructure. Used by application builders, researchers, and businesses needing access to state-of-the-art LLMs.
openai api · anthropic api · gemini api · xai api · mistral ai · cohere · deepseek api · moonshot ai
Platforms that host and serve open-source large language models via APIs, allowing developers and enterprises to run models like Llama, Mistral, or Qwen without building inference infrastructure. Used by application builders, researchers, and organizations avoiding proprietary model lock-in.
together ai · fireworks ai · deepinfra · openrouter · hugging face inference providers · baseten · novita ai · hyperbolic
Custom silicon inference refers to specialized hardware accelerators (TPUs, custom ASICs, neuromorphic chips) designed by companies to run inference workloads. Machine learning teams, cloud providers, and AI labs use these to reduce latency and cost per inference compared to general-purpose GPUs.
groq · cerebras · sambanova
Managed platforms that run inference on hosted large language models without users managing infrastructure or containers. Teams building applications, researchers, and enterprises use these to call models via APIs.
amazon bedrock · azure ai foundry · google vertex ai · oracle oci generative ai · ibm watsonx.ai · databricks mosaic ai
Serverless GPU inference provides on-demand GPU compute for running language model inference without managing infrastructure; used by application developers, API providers, and enterprises running LLM workloads at variable scale.
modal · replicate · runpod · fal · coreweave