
26M-parameter open-source function-call model for small consumer devices.
Needle is a 26-million-parameter function-call model distilled from Gemini into a Simple Attention Network. It is designed to run on phones, watches, glasses, and other low-power devices, achieving 6,000 tokens per second prefill and 1,200 tokens per second decode when running on the Cactus engine.
Weights, training code, and a fine-tuning playground are open-sourced on GitHub and Hugging Face. The model targets single-shot tool use and retrieval-and-assembly tasks rather than open-ended reasoning.


Google Gemini is Google’s AI assistant for chat, search, writing, coding, and creation.
Google Gemini is Google’s AI assistant for chat, search, writing, coding, and creation.



Conifer Gateway is one OpenAI- and Anthropic-compatible endpoint for more than 200 models.



Pre-optimized APIs for Kimi, DeepSeek, GLM on Baseten Inference Stack.


LangChain middleware that adds ScaleDown small language models to agents for summarization, compression and extraction.





Open source AI gateway routing every model behind one zero-markup OpenAI-compatible endpoint.


Models.dev is an open-source directory and API for discovering and comparing AI models.


Hosted gateway for routing and governing traffic across hosted and self-hosted AI models.


A unified OpenAI-compatible API and gateway for accessing hundreds of LLMs from many providers.




Baseten Inference Stack for high performance with custom kernels, decoding, caching on optimized infra.


Train models on B200s, T4s and inference-optimized infra with one-click deploy to production.


GLM Model API is Z.ai's developer platform for calling its GLM large language models and agent services over an API.


OpenAI-compatible API proxy with built-in conversational memory supporting 300+ models.



Pip is a marketing-native AI workspace for marketing stacks with marketing-trained AI agents and human engineering support.
Conifer Gateway is one OpenAI- and Anthropic-compatible endpoint for more than 200 models.

Fugu is a multi-agent orchestration API that routes complex tasks across a pool of frontier language models.

Pre-optimized APIs for Kimi, DeepSeek, GLM on Baseten Inference Stack.

LangChain middleware that adds ScaleDown small language models to agents for summarization, compression and extraction.

River API lets developers fine-tune and run reinforcement learning on open-weight models.

Custom AI model builder for enterprise infrastructure.

Adaptive OpenAI-compatible endpoint coordinating open-weight models for chat and agents.

Open source AI gateway routing every model behind one zero-markup OpenAI-compatible endpoint.

Models.dev is an open-source directory and API for discovering and comparing AI models.

Hosted gateway for routing and governing traffic across hosted and self-hosted AI models.

A unified OpenAI-compatible API and gateway for accessing hundreds of LLMs from many providers.

Stable is a slash-command layer that brings any model into an existing coding agent.

LLM-powered AI agent for lipid nanoparticle research and formulation optimization.

Baseten Inference Stack for high performance with custom kernels, decoding, caching on optimized infra.

Train models on B200s, T4s and inference-optimized infra with one-click deploy to production.

GLM Model API is Z.ai's developer platform for calling its GLM large language models and agent services over an API.

OpenAI-compatible API proxy with built-in conversational memory supporting 300+ models.

Managed GPU cloud for inference, fine-tuning, and prompt workflows.

Pip is a marketing-native AI workspace for marketing stacks with marketing-trained AI agents and human engineering support.
