Fournisseurs

Les laboratoires, clouds et routeurs dont les modèles figurent au catalogue. Les prix affichés sont ceux d’OPTES, pas les leurs.

OPTES

OPTESOPTES's own model family: text, code, vision, images, video, voice, music and search under one key.12 modèles

Laboratoires de modèles

AI21 LabsThe Jamba family of long-context models.2 modèles AnthropicThe Claude family (Fable, Opus, Sonnet, Haiku) through the Claude API.4 modèles CohereCanadian lab focused on enterprise models; Aya Expanse is its multilingual line.1 modèles Core42 CompassG42's model platform in the UAE, home of the Arabic-first Seraj model.5 modèles Google AI StudioThe Gemini family through the Gemini API (Google AI Studio).6 modèles MetaMeta's Model API (public preview) with the Muse family; Llama models run on other providers.4 modèles Mistral AIFrench lab: Mistral Medium, Large and Small, Codestral, Ministral and embeddings.6 modèles OpenAIThe GPT-6 and GPT-5.x families through OpenAI's own API platform.4 modèles PerplexitySonar, a search-grounded model, plus selected third-party models through its Agent API.3 modèles Reka AIMultimodal models that read text, images, video and audio.3 modèles UpstageKorean lab: the Solar models and document AI.3 modèles WriterEnterprise lab behind the Palmyra models with 1M-token context.3 modèles xAIThe Grok family, including Grok Build for code and Grok Imagine for images.4 modèles

Laboratoires chinois, plateformes internationales

Alibaba Cloud Model StudioThe Qwen family through Alibaba Cloud's international Model Studio.3 modèles BytePlus ModelArkByteDance's international platform: the Seed models and hosted DeepSeek and GLM.3 modèles DeepSeekDeepSeek's own platform: V4 Pro and V4.1 Flash, with 1M context and a 50% off-peak discount.2 modèles MiniMaxMiniMax's international platform: the M-series text models with 1M context.2 modèles Moonshot AIThe Kimi models through Moonshot's international platform.4 modèles StepFunStepFun's international platform: Step 5 and the fast Step 3.x Flash models.3 modèles Tencent Cloud TokenHubTencent's international hub: the Hy models plus Kimi, GLM, DeepSeek and MiniMax.3 modèles Xiaomi MiMoXiaomi's MiMo models, priced low outside China.2 modèles Z.aiThe GLM family through Z.ai's international platform.5 modèles

Plateformes cloud

Amazon BedrockAWS's model service: Claude, GPT-6 Astra, Amazon Nova, Llama, Mistral, DeepSeek, Kimi and Grok.13 modèles Google Cloud Vertex AIGoogle Cloud's model platform: Gemini plus Claude, Llama and Mistral on one bill.7 modèles IBM watsonx.aiIBM's model platform: Granite plus Llama, Mistral and gpt-oss.5 modèles Microsoft Azure AI FoundryMicrosoft's model platform: OpenAI models plus DeepSeek, Grok, Kimi, Llama and Mistral.11 modèles Oracle OCI Generative AIOracle Cloud's model service, with regions in Dubai, Abu Dhabi and Riyadh.5 modèles

Fournisseurs d’inférence

AkashMLOpen models on the decentralised Akash GPU network.7 modèles BasetenAmong the fastest providers for gpt-oss-120b and Kimi K3, with near-lowest prices.8 modèles CerebrasThe fastest published speeds in the research: about 3,000 tokens a second on gpt-oss-120b.2 modèles ChutesDecentralised inference where every model runs inside a trusted execution environment (TEE).9 modèles Cloudflare Workers AIModels served on Cloudflare's network, priced per 1M tokens from its Neuron units.13 modèles CoreWeaveGPU cloud with serverless inference (formerly W&B Inference) and a very recent catalog.14 modèles CrusoeManaged inference with low cached-input prices: DeepSeek, Kimi, GLM, gpt-oss and Nemotron.9 modèles DeepInfraAmong the lowest prices for gpt-oss, GLM-5.3, DeepSeek V4 Flash and MiniMax M3.15 modèles falA marketplace of image, video and audio generation models.3 modèles FeatherlessMore than 4,000 open models and fine-tunes from Hugging Face behind one API.12 modèles Fireworks AIFast serverless inference with Standard and Fast tiers for the big open models.7 modèles FriendliAIFast inference engine; a focused serverless list around GLM-5.x, DeepSeek and MiniMax.8 modèles GMI CloudGPU cloud with an inference engine for DeepSeek, Kimi, GLM, Qwen and MiniMax.11 modèles GroqVery fast inference on its own LPU chips; a small public catalog.6 modèles inference.netInference platform with open models and its own Schematron extraction models.10 modèles io.netDecentralised GPU network with IO Intelligence, an API for open models.12 modèles ModalServerless GPU platform with a shared per-token API for Kimi K3, GLM-5.3 and more.4 modèles Nebius Token FactoryEuropean inference cloud that publishes a measured speed for every model.14 modèles Novita AIA broad catalog of 120 open models for text, images, video and audio.17 modèles NscaleBritish AI cloud with serverless endpoints for gpt-oss, Llama and Qwen.7 modèles OVHcloud AI EndpointsEuropean serverless endpoints for gpt-oss, Llama 3.3 and Qwen.6 modèles ParasailDay-zero support for big open models and some of the lowest DeepSeek and gpt-oss prices.17 modèles ReplicateA large catalog of image, video and audio models; a few language models.2 modèles RunpodGPU cloud with public endpoints for Kimi models, including Kimi K3 with 1M context.4 modèles SambaNovaInference on SambaNova's own chips: DeepSeek, Llama, gpt-oss, MiniMax and Gemma.7 modèles SiliconFlowInternational platform with some of the lowest prices for DeepSeek V4.1 Flash and gpt-oss.14 modèles Together AIOne of the widest catalogs of open-weights models: DeepSeek, Kimi, GLM, Qwen, Llama and more.12 modèles

Routeurs et passerelles

302.AIA gateway to Western and Chinese models plus image, video and audio generation.1 modèles AIHubMixA gateway to text, image, video and audio models from the big labs.4 modèles Eden AIOne API over 500+ models, including specialised expert models.3 modèles Hugging Face Inference ProvidersHugging Face's router over partner providers, at the partners' prices.10 modèles OpenRouterA router over 500+ models from 80+ providers.3 modèles RequestyA router over 600+ models from 20+ providers.3 modèles TokenGOSelf-operated open models: DeepSeek, Kimi, GLM, Qwen and MiniMax.4 modèles