Find the right AI model for any task

Background

DeepInfra runs popular open-source models (Llama, Mistral, DeepSeek, Qwen, etc.) on managed GPU clusters and exposes them through a drop-in-compatible OpenAI REST API. Developers pay only for tokens generated — there is no server allocation or idle cost. The /v1/models endpoint is publicly accessible without a token, making it easy to enumerate available models.

Mission: Make open-source AI inference affordable and highly available for every developer.

Profile

Region

North America

Country

United States

Headquarters

San Francisco, California, USA

Founded

2022

Pricing

Pay-per-token; no minimum commitment; free tier for low-volume use.

Background

Hugging Face hosts the world's largest public repository of machine-learning models and datasets, anchored by the Transformers library used by millions of researchers. Its Inference API lets developers call models over HTTPS without managing GPU infrastructure. HF also operates Spaces (shareable ML demos), the Hub (version control for model weights), and enterprise offerings including dedicated inference endpoints and private model repositories.

Mission: Democratise good machine learning for researchers and practitioners everywhere.

Profile

Region

North America

Country

United States

Headquarters

New York, NY, USA

Founded

2016

Pricing

Free tier for public models; pay-per-compute for Inference Endpoints; enterprise contracts.

Background

OpenAI develops foundation models for text, images, audio, and video (GPT, DALL-E, Whisper, Sora) and exposes them through a platform API used by millions of developers. The company also offers ChatGPT, a consumer product, and enterprise agreements for organisations. Research areas include alignment, interpretability, and scalable reinforcement learning from human feedback.

Mission: Ensure that artificial general intelligence benefits all of humanity.

Profile

Region

North America

Country

United States

Headquarters

San Francisco, California, USA

Founded

2015

Pricing

Pay-per-token for API; ChatGPT Plus/Pro subscription for consumer.

Background

Ollama bundles model weights, quantisation, and a local REST server into a single desktop application. Users run `ollama pull llama3` and immediately get a local API on localhost:11434 compatible with the OpenAI chat format. The Ollama Library (ollama.com/library) is a curated registry of ready-to-use model packages, including popular open-weight models from Meta, Mistral, Google, and many others.

Mission: Make running powerful AI models locally as easy as running any other software.

Profile

Region

North America

Country

United States

Headquarters

San Francisco, California, USA

Founded

2023

Pricing

Free and open source (MIT + Apache licences).

Background

OVHcloud is France's largest cloud provider and Europe's second largest, operating 43+ datacentres worldwide. Its AI Endpoints service runs popular open-weight models (Llama, Mistral, Stable Diffusion, etc.) on NVIDIA-GPU-backed infrastructure in France, providing an OpenAI-compatible REST API for EU organisations that must keep data in Europe. OVHcloud also provides Object Storage, Managed Kubernetes, Bare-Metal, and Private Cloud.

Mission: Provide independent, trustworthy, and sustainable cloud infrastructure anchored in European values.

Profile

Region

Europe

Country

France

Headquarters

Roubaix, France

Founded

1999

Pricing

Pay-per-token or metered compute; free tier up to monthly quota.

Background

Anthropic focuses on frontier AI safety research alongside building commercial products. Its Claude models (Haiku, Sonnet, Opus) are available via API and through the Claude.ai consumer app. The company publishes research on Constitutional AI, RLHF, scalable oversight, and mechanistic interpretability. Anthropic distributes Claude through Amazon Bedrock and Google Cloud Vertex AI as well as its own API.

Mission: Build AI systems people can rely on while driving a race to the top on safety.

Profile

Region

North America

Country

United States

Headquarters

San Francisco, California, USA

Founded

2021

Pricing

Pay-per-token (input + output); Claude Pro subscription for consumer.

Background

DeepSeek is an AI research division of High-Flyer Capital Management, a Hangzhou-based quantitative fund. Their models — including DeepSeek-V3 and DeepSeek-R1 — achieved top-tier benchmark performance at a fraction of the compute cost reported by comparable US models, demonstrating that efficient training techniques could close the frontier gap. DeepSeek releases model weights publicly under permissive licences, and their inference API offers highly competitive per-token pricing.

Mission: Solve AGI through open, efficient, and principled research that benefits all of humanity.

Profile

Region

Asia-Pacific

Country

China

Headquarters

Hangzhou, Zhejiang, China

Founded

2023

Pricing

Pay-per-token; among the lowest prices globally for frontier model quality.

Background

MiniMax builds large multimodal foundation models targeting text, speech, image, and video generation. Their MiniMax-Text-01 and MiniMax-01-VL are available as API and open-weight. Consumer products include Hailuo AI (video generation), Talkie (AI companion), and Xinghuo (enterprise assistant). MiniMax-01 (released January 2025 as open-weight under MIT licence) attracted global attention with a 4 million-token context window.

Mission: Build widely accessible general-purpose AI that enriches everyday life and empowers enterprises.

Profile

Region

Asia-Pacific

Country

China

Headquarters

Shanghai, China

Founded

2021

Pricing

Pay-per-token for text; per-video for Hailuo.

Background

OpenRouter provides a single OpenAI-compatible API endpoint that transparently routes model requests to the most appropriate or cheapest upstream provider. Developers integrate once and gain access to hundreds of models without managing separate API keys for every lab. OpenRouter also surfaces per-token pricing comparisons, model context windows, and latency benchmarks in real time.

Mission: Make large language models more accessible and comparable through a unified API.

Profile

Region

North America

Country

United States

Headquarters

San Francisco, California, USA

Founded

2023

Pricing

Pass-through per-token pricing with a small margin; free tier available with rate limits.

Background

Cohere builds large language models purpose-built for enterprise deployments, with a focus on retrieval-augmented generation (RAG), document search, and safe generation in regulated industries. Their product line includes Command (text generation + tool calling), Embed (multilingual embeddings for semantic search), Rerank (cross-encoder reranking to boost retrieval quality), and North (an enterprise AI agent platform). Cohere offers on-premise and private-cloud deployment via its Model Vault, important for regulated sectors.

Mission: Empower businesses with AI that enhances human wellbeing and drives productivity that reaches everyone.

Profile

Region

North America

Country

Canada

Headquarters

Toronto, Ontario, Canada

Founded

2019

Pricing

Pay-per-token; enterprise SaaS agreements; private deployment pricing on request.

Background

AI21 Labs develops foundation models and enterprise AI platforms with a focus on reliability and production readiness. Their Jamba models are unique in combining State Space Model (SSM / Mamba) layers with traditional Transformer attention, yielding high throughput on very long contexts. The company also offers Maestro, an AI orchestration platform for multi-step agentic workflows, and targets sectors including finance, healthcare, tech, and defence.

Mission: Provide trustworthy artificial intelligence that powers humanity toward superproductivity.

Profile

Region

Middle East

Country

Israel

Headquarters

Tel Aviv, Israel

Founded

2017

Pricing

Pay-per-token; free-tier trial credits on sign-up.

Background

StepFun (阶跃星辰) develops frontier foundation models with a focus on multimodal capabilities (image understanding, long video analysis) and strong Chinese-language performance. Their Step-2 model achieved competitive results in mathematics and reasoning benchmarks. The platform API supports step-1, step-1v (vision), and step-2 model variants. The company has attracted significant backing from investors in China's $10B AI race.

Mission: Build frontier multimodal AI that advances human-level understanding and reasoning.

Profile

Region

Asia-Pacific

Country

China

Headquarters

Shanghai, China

Founded

2023

Pricing

Pay-per-token.

Background

Alibaba Cloud is the cloud-computing arm of Alibaba Group, and its AI effort centres on the Qwen (通义千问) series — a family of text, visual (Qwen-VL), audio (Qwen-Audio), and code models. The DashScope / Bailian API provides OpenAI-compatible access to Qwen models at competitive pricing. Qwen2.5 models, released in late 2024, achieved state-of-the-art results on many coding and reasoning benchmarks and are among the most downloaded open-weight models globally.

Mission: Provide AI infrastructure and world-class models that power global digital transformation.