OpenRouter's October data recap reports 790 trillion tokens processed in the month, a 25x increase from January, alongside metrics on agentic usage, token pricing, model lab retention and inference providers. The figures offer one of the few public quantitative views of real inference demand growth and the API market landscape.
Inference Provider
推理服务商
Third-party cloud platforms providing compute resources and APIs for running AI models.
Curated Intelligence
Baseten is now a supported Inference Provider on the Hugging Face Hub, allowing users to access Baseten-hosted models via standard HF interfaces. The integration passes through provider costs without markup, and PRO users receive $2 in monthly credits usable across all providers.
DeepInfra is now a supported Inference Provider on the Hugging Face Hub, enabling direct access to hosted models like Kimi-K2.6 and GLM-5.1. Text generation is currently available, with multimodal capabilities rolling out soon.
Public AI announced that its models are now available via Hugging Face Inference Providers. The update confirms platform integration but lacks details on specific model versions, benchmarks, or pricing.
Groq has officially joined Hugging Face Inference Providers, offering low-latency model inference powered by its LPU architecture. This integration allows developers to access Groq's high-performance compute resources via a unified API interface.
Featherless AI announces its models are now available via Hugging Face Inference Providers. This integration allows developers to call hosted Featherless models directly, though the source text lacks details on supported model lists, performance metrics, or pricing.
Hugging Face Hub announced the integration of three new serverless inference providers: Hyperbolic, Nebius AI Studio, and Novita. This expansion offers developers more compute options and cost flexibility for model deployment, with PRO users receiving $2 in monthly credits usable across all providers.
Hugging Face Hub officially integrates Fireworks.ai into its Inference Providers ecosystem, enabling users to access high-speed serverless inference directly from model pages. This update simplifies running LLMs within the HF ecosystem, with PRO users receiving $2 in monthly credits usable across all providers.