Diese Seite ist nur auf Englisch verfügbar. Zur englischen Version wechseln →
Wir arbeiten daran, weitere Seiten zu übersetzen. Danke für Ihr Verständnis!
AI Companies & Research Labs
AI Labs
32 providers building the models you use every day. Explore their full catalog, compare pricing, and start chatting.
OpenAI
51 models
Creator of GPT-4o, o1, and DALL-E. Pioneered large-scale language models and RLHF alignment.
Qwen (Alibaba)
39 models
Alibaba Cloud's model team. Builds the Qwen series covering text, vision, code, and audio.
Mistral AI
27 models
European AI lab building efficient open-weight models. Known for Mixtral MoE architecture and multilingual strength.
Google DeepMind
19 models
Builds Gemini and PaLM models at Google DeepMind. Leads in multimodal AI and long-context research.
Meta
15 models
Releases the open-weight Llama model family. Largest contributor to open-source foundation models.
Anthropic
12 models
Safety-focused lab behind the Claude model family. Known for Constitutional AI and long-context capabilities.
DeepSeek
10 models
Chinese AI lab producing high-performance open-weight models. Strong in coding and mathematical reasoning.
xAI
8 models
Elon Musk's AI company building Grok. Trained on real-time data with a focus on truthfulness and wit.
Zhipu AI
8 models
Chinese AI company behind the GLM model family. Strong in bilingual Chinese-English tasks and tool use.
Allen Institute for AI
7 models
Non-profit research institute building truly open AI. Known for OLMo and Tulu models with fully released training data, code, and weights.
MiniMax
6 models
Chinese AI startup building multimodal models for content creation, voice, and video generation.
Amazon
5 models
Develops the Nova model family for AWS Bedrock. Optimized for enterprise workloads and AWS integration.
Perplexity
5 models
AI search company with models tuned for real-time information retrieval and citation-grounded answers.
Nous Research
5 models
Open-source AI research lab known for the Hermes model family. Specializes in reasoning, function calling, and hybrid thinking modes.
Baidu
5 models
Chinese tech giant behind the ERNIE model family. Operates Wenxin Yiyan (ERNIE Bot), one of China's most widely used AI assistants.
Cohere
4 models
Enterprise-focused lab specializing in RAG, embeddings, and multilingual models for business use cases.
Moonshot AI
4 models
Chinese AI company behind Kimi. Specializes in ultra-long context windows and document understanding.
Arcee AI
4 models
Open-intelligence lab building efficient foundation models for edge, on-prem, and cloud. Pioneers model merging techniques.
Aion Labs
4 models
AI lab building powerful reasoning models via API. Known for Aion-1.0, a multi-model reasoning system using Tree of Thoughts and MoE.
NVIDIA
3 models
Builds Nemotron models optimized for NVIDIA hardware. Leading GPU maker powering most AI training.
ByteDance Seed
3 models
ByteDance's AI research division behind Seedream and Doubao models. Builds multimodal and generative AI for TikTok's parent company.
Microsoft
2 models
Develops the Phi series of small efficient models and MAI for enterprise. Major investor in OpenAI.
Inflection AI
2 models
Built Pi, a personal AI assistant. Focus on emotionally intelligent and conversational AI systems.
Liquid AI
2 models
MIT CSAIL spin-off building Liquid Foundation Models, a novel architecture with reduced memory requirements designed to work at every scale.
Inception
2 models
AI lab pioneering diffusion-based large language models. Known for Mercury, the first commercial-scale diffusion LLM with 5-10x faster generation.
AI21 Labs
1 model
Builds the Jamba architecture combining Mamba SSM with Transformers. Focus on enterprise NLP applications.
Writer
1 model
Enterprise AI platform with proprietary Palmyra models built for business writing and content workflows.
Deep Cogito
1 model
European AI lab building the Cogito model family. Combines hybrid training with reasoning and multilingual capabilities.
EleutherAI
1 model
Non-profit open-source AI research institute. Created GPT-J, GPT-NeoX, the Pythia model suite, and The Pile dataset.
Xiaomi
1 model
Chinese tech giant whose AI team builds the MiMo family of open-source language models with strong math and code reasoning.
IBM Granite
1 model
IBM Research's open-source Granite model family built for enterprise. Strong in code generation, summarization, and RAG applications.
Tencent
1 model
Chinese tech giant behind the Hunyuan model series covering text, image, video, and multimodal generation across Tencent's ecosystem.
The AI Labs Building Today's Models
Understanding who builds the models helps you pick the right one.
Frontier labs
OpenAI (GPT-4o, o1), Anthropic (Claude), and Google DeepMind (Gemini) lead on general-purpose benchmarks and multimodal capabilities. They invest billions in training runs and offer managed APIs with built-in safety layers.
Open-source leaders
Meta (Llama), Mistral AI, DeepSeek, and Alibaba (Qwen) release open-weight models you can self-host, fine-tune, and inspect. These labs have narrowed the gap with proprietary models while enabling full control over deployment and data privacy.
Specialists & newcomers
Focused labs like Cohere (enterprise RAG), AI21 (Jamba architecture), and xAI (Grok) target specific niches. Newer entrants regularly push the state of the art in coding, math, or multilingual tasks at lower cost.
Why provider choice matters
Each lab has different strengths, pricing structures, data policies, and rate limits. Comparing models within and across providers helps you find the best balance of quality, cost, and compliance for your use case.
Provider data is sourced from OpenRouter. Model counts reflect actively available models. Click any lab to explore their full model catalog with pricing, benchmarks, and instant chat.