# AI Gateway Directory

> Labs, inference platforms, gateways and observability for LLM apps.

## Frontier labs
- [Anthropic API](https://model-desk.pages.dev/tool/anthropic-api/): Claude models through the Anthropic developer platform
- [Google Gemini API](https://model-desk.pages.dev/tool/google-gemini-api/): Gemini models via Google AI Studio and the Gemini API
- [OpenAI API](https://model-desk.pages.dev/tool/openai-api/): GPT and reasoning models behind one developer API
- [Alibaba Cloud Model Studio](https://model-desk.pages.dev/tool/alibaba-cloud-model-studio/): Qwen models and a model service on Alibaba Cloud
- [Amazon Nova](https://model-desk.pages.dev/tool/amazon-nova/): Amazon's own foundation models on Bedrock
- [Cohere](https://model-desk.pages.dev/tool/cohere/): Enterprise language models for search and agents
- [DeepSeek API](https://model-desk.pages.dev/tool/deepseek-api/): Low-cost reasoning and chat models, OpenAI-compatible
- [Mistral AI](https://model-desk.pages.dev/tool/mistral-ai/): French lab with hosted APIs and open models
- [Perplexity API](https://model-desk.pages.dev/tool/perplexity-api/): Search-grounded answers and a web search API
- [xAI API](https://model-desk.pages.dev/tool/xai-api/): Grok models for developers

## Open-weight providers
- [Llama](https://model-desk.pages.dev/tool/llama/): Meta's open-weight model family
- [DeepSeek-V3](https://model-desk.pages.dev/tool/deepseek-v3/): Open weights for DeepSeek's V3 mixture-of-experts model
- [Gemma](https://model-desk.pages.dev/tool/gemma/): Google's lightweight open models
- [GLM-4.5](https://model-desk.pages.dev/tool/glm-4-5/): Z.ai's open GLM model series
- [gpt-oss](https://model-desk.pages.dev/tool/gpt-oss/): OpenAI's open-weight reasoning models
- [IBM Granite](https://model-desk.pages.dev/tool/ibm-granite/): IBM's open enterprise language models
- [Kimi K2](https://model-desk.pages.dev/tool/kimi-k2/): Moonshot AI's open mixture-of-experts model
- [NVIDIA Nemotron](https://model-desk.pages.dev/tool/nvidia-nemotron/): Open NVIDIA models with open data and recipes
- [OLMo](https://model-desk.pages.dev/tool/olmo/): Fully open language models from Ai2
- [Phi](https://model-desk.pages.dev/tool/phi/): Microsoft's small language models
- [Qwen3](https://model-desk.pages.dev/tool/qwen3/): Alibaba's open Qwen3 language models

## Inference platforms
- [Ollama](https://model-desk.pages.dev/tool/ollama/): Run open models locally with one command
- [Amazon Bedrock](https://model-desk.pages.dev/tool/amazon-bedrock/): Managed access to many foundation models on AWS
- [Azure AI Foundry](https://model-desk.pages.dev/tool/azure-ai-foundry/): Microsoft's platform for building with models and agents
- [Baseten](https://model-desk.pages.dev/tool/baseten/): Production inference for open and custom models
- [Cerebras](https://model-desk.pages.dev/tool/cerebras/): Inference on wafer-scale chips
- [Cloudflare Workers AI](https://model-desk.pages.dev/tool/cloudflare-workers-ai/): Run models on Cloudflare's network
- [DeepInfra](https://model-desk.pages.dev/tool/deepinfra/): Pay-per-token API for open models
- [Fireworks AI](https://model-desk.pages.dev/tool/fireworks-ai/): Fast inference for open and custom models
- [Google Vertex AI](https://model-desk.pages.dev/tool/google-vertex-ai/): Google Cloud's platform for models and ML
- [Groq](https://model-desk.pages.dev/tool/groq/): Inference on custom LPU chips
- [Hugging Face Inference Providers](https://model-desk.pages.dev/tool/hugging-face-inference-providers/): One API across many inference partners
- [llama.cpp](https://model-desk.pages.dev/tool/llama-cpp/): C/C++ inference for GGUF models on everyday hardware
- [Modal](https://model-desk.pages.dev/tool/modal/): Serverless GPUs defined in Python
- [Replicate](https://model-desk.pages.dev/tool/replicate/): Run models through an API without managing servers
- [SambaNova](https://model-desk.pages.dev/tool/sambanova/): Inference cloud on custom RDUs
- [Together AI](https://model-desk.pages.dev/tool/together-ai/): Serverless and dedicated inference for open models
- [vLLM](https://model-desk.pages.dev/tool/vllm/): High-throughput open-source LLM serving engine

## Gateways & routers
- [OpenRouter](https://model-desk.pages.dev/tool/openrouter/): One API and one bill for hundreds of models
- [Bifrost](https://model-desk.pages.dev/tool/bifrost/): Fast open-source LLM gateway from Maxim
- [Cloudflare AI Gateway](https://model-desk.pages.dev/tool/cloudflare-ai-gateway/): Proxy for AI traffic with caching and rate limits
- [Eden AI](https://model-desk.pages.dev/tool/eden-ai/): European AI gateway with one API for many models
- [Kong AI Gateway](https://model-desk.pages.dev/tool/kong-ai-gateway/): AI traffic management on the Kong API gateway
- [LiteLLM](https://model-desk.pages.dev/tool/litellm/): Open-source proxy and SDK for 100+ LLM providers
- [Portkey](https://model-desk.pages.dev/tool/portkey/): AI gateway with guardrails, caching and observability
- [Requesty](https://model-desk.pages.dev/tool/requesty/): LLM router with failover and governance
- [TensorZero](https://model-desk.pages.dev/tool/tensorzero/): Open-source LLM gateway, evals and optimization
- [Vercel AI Gateway](https://model-desk.pages.dev/tool/vercel-ai-gateway/): One endpoint for models from many providers

## Observability
- [Arize Phoenix](https://model-desk.pages.dev/tool/arize-phoenix/): AI observability and evaluation you can self-host
- [Braintrust](https://model-desk.pages.dev/tool/braintrust/): Evals, logging and prompt tooling for AI products
- [Datadog LLM Observability](https://model-desk.pages.dev/tool/datadog-llm-observability/): LLM traces inside Datadog
- [Helicone](https://model-desk.pages.dev/tool/helicone/): Open-source LLM observability through a proxy or SDK
- [Langfuse](https://model-desk.pages.dev/tool/langfuse/): Open-source LLM observability, evals and prompt management
- [LangSmith](https://model-desk.pages.dev/tool/langsmith/): Tracing, evals and deployment from the LangChain team
- [OpenLLMetry](https://model-desk.pages.dev/tool/openllmetry/): OpenTelemetry instrumentation for LLM apps
- [Opik](https://model-desk.pages.dev/tool/opik/): Open-source LLM tracing and evals from Comet
- [PostHog LLM analytics](https://model-desk.pages.dev/tool/posthog-llm-analytics/): LLM traces next to product analytics
- [Pydantic Logfire](https://model-desk.pages.dev/tool/pydantic-logfire/): OpenTelemetry observability from the Pydantic team
- [Weights & Biases Weave](https://model-desk.pages.dev/tool/weights-biases-weave/): Tracing and evaluation for AI applications

## Fine-tuning
- [Axolotl](https://model-desk.pages.dev/tool/axolotl/): Config-driven fine-tuning for open LLMs
- [Hugging Face TRL](https://model-desk.pages.dev/tool/hugging-face-trl/): Train language models with SFT, DPO and RL
- [LLaMA-Factory](https://model-desk.pages.dev/tool/llama-factory/): Unified fine-tuning for 100+ language models
- [OpenAI fine-tuning](https://model-desk.pages.dev/tool/openai-fine-tuning/): Fine-tune OpenAI models on your own examples
- [Oumi](https://model-desk.pages.dev/tool/oumi/): Open platform to train, evaluate and deploy models
- [Tinker](https://model-desk.pages.dev/tool/tinker/): Fine-tuning API for open-weight models
- [torchtune](https://model-desk.pages.dev/tool/torchtune/): PyTorch-native library for fine-tuning LLMs
- [Unsloth](https://model-desk.pages.dev/tool/unsloth/): Faster, lower-memory fine-tuning of open models

## Embeddings & rerank
- [BGE (FlagEmbedding)](https://model-desk.pages.dev/tool/bge-flagembedding/): Open BGE embedding and reranker models
- [Cohere Embed](https://model-desk.pages.dev/tool/cohere-embed/): Multilingual and multimodal embeddings
- [Cohere Rerank](https://model-desk.pages.dev/tool/cohere-rerank/): Rerank search results with a hosted model
- [Gemini embeddings](https://model-desk.pages.dev/tool/gemini-embeddings/): Embedding models in the Gemini API
- [Jina AI](https://model-desk.pages.dev/tool/jina-ai/): Embeddings, rerankers and web reader APIs
- [Mixedbread](https://model-desk.pages.dev/tool/mixedbread/): Search, embedding and rerank models and APIs
- [OpenAI embeddings](https://model-desk.pages.dev/tool/openai-embeddings/): Text embedding models from OpenAI
- [Qwen3 Embedding](https://model-desk.pages.dev/tool/qwen3-embedding/): Open embedding and reranking models from the Qwen team
- [Sentence Transformers](https://model-desk.pages.dev/tool/sentence-transformers/): Python library for embedding and rerank models
- [Text Embeddings Inference](https://model-desk.pages.dev/tool/text-embeddings-inference/): Hugging Face server for embedding and rerank models
- [Voyage AI](https://model-desk.pages.dev/tool/voyage-ai/): Embedding and reranking models for retrieval