Dedicated

Deploy any open-weight model on GPUs reserved for you

283 models from 64 makers, ready to request: chat and reasoning, coding, vision, OCR, embeddings, rerankers, speech, and text-to-speech. Private endpoint, European GPUs, billed per GPU-hour. Don't see yours? Ask anyway, fine-tunes included.

Request a deployment ↓All 283 models/dedicated.md

Reserved capacity

GPUs reserved for your deployment: predictable latency and guaranteed throughput, no noisy neighbours.

European GPUs

Every request processed on European GPUs, with the same request-level logging and audit trail as serverless.

Same API

Private endpoint with the OpenAI-compatible API you already use. Moving a hot path over is a routing change.

Priced per GPU-hour

Monthly commitments sized to your throughput and latency targets, with no shared-pool contention.

Most popular

ModelParams
DeepSeek-V4.1-FlashserverlesstrendingvisionMoE
DeepSeek		763B
GLM-5.3-FlashserverlesstrendingvisionMoE
Z.ai		321B
GLM-5.3serverlesstrendingMoE
Z.ai		753B
DeepSeek-V4-Flash-0731serverlessMoE
DeepSeek		304B
MiniMax-M3.1-FlashnewtrendingvisionMoE
MiniMax		220B
Qwen3.8-27Btrendingvision
Qwen		28B
Kimi-K3serverlesstrendingvisionMoE
Moonshot AI		2.78T
Qwen3.8-Flash-NexttrendingvisionMoE
Qwen		180B
Kokoro-82Mtrending
hexgrad		—
DeepSeek-V4-FlashserverlessMoE
DeepSeek		291B
gemma-4-31B-ittrendingvision
Google		31B
speaker-diarization-community-1trending
pyannote		—

Open a model for its footprint, license summary, and FAQ. Models marked serverless are on the per-token catalog too. Browse all 283 models →

Browse by maker · 64

Qwen56Google19NVIDIA17DeepSeek16Z.ai16Mistral AI12Moonshot AI10Tencent8MiniMax7Microsoft7OpenBMB6IBM6OpenAI5Xiaomi MiMo5Meta Llama5Liquid AI5Jina AI5inclusionAI5Baidu4Cohere4StepFun4Alibaba NLP4BAAI3ByteDance Seed3Meta Muse2Resemble AI2Thinking Machines2poolside2Hugging Face2PaddlePaddle2Nomic2Upstage2Moondream2Swiss AI Initiative2Meituan LongCat2hexgrad1pyannote1Fish Audio1Datalab1Aleph Alpha1XingChen AGI1LightOn1Sesame1FunAudioLLM1dots.ocr1Nari Labs1intfloat1Reflection AI1Kwaipilot1Snowflake1Canopy Labs1IndexTeam1Perplexity1Nanonets1LG AI Research1Reducto1Kyutai1ServiceNow1ZeroEntropy1OpenDataLab1Ai21Mixedbread1InternLM1EuroLLM1

Request a deployment

Tell us the model and roughly how much traffic you expect. Fine-tunes and distillations of these models work the same way; see custom models.

  1. 01
    We size it

    A performance tier, the capacity behind it, and a quote for your volume.

  2. 02
    We benchmark it

    Bring one workload with a sample of real traffic; we come back with latency and throughput numbers.

  3. 03
    It goes live

    We provision the deployment. You call it with the same API and model id as before.

No commitment. Prefer email? [email protected]