| AI21 | official | unknown — General OpenAI API compatibility has not been verified from the official documentation for this profile. | Usage-based pricing depends on the selected AI21 model and API. First-party published pricing; not a resale markup. | Available models and task-specific APIs are listed in AI21's documentation. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Alibaba Cloud Model Studio | official | partial — Alibaba Cloud documents OpenAI-compatible access for supported models through the documented endpoint; native DashScope features and modes differ. | Usage-based pricing depends on model, modality, deployment, and Alibaba Cloud billing conditions. Alibaba Cloud publishes platform pricing; no universal markup is inferred. | The Model Studio catalog includes Qwen and other supported models; availability depends on API and region. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Anthropic | official | partial — Anthropic documents an OpenAI SDK compatibility endpoint, but API features are not identical. | Usage-based token pricing, with rates varying by Claude model. First-party published API pricing; not a resale markup. | Claude model names, capabilities, and lifecycle are listed in the official documentation. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| AWS Bedrock | official | unknown — No general OpenAI-compatible endpoint is asserted here; Bedrock uses AWS APIs and model-specific request schemas. | Model- and usage-dependent pricing, with on-demand and provisioned options for supported models. AWS publishes Bedrock prices by model and usage mode; no uniform markup is inferred. | The catalog contains models from multiple providers; access and availability vary by model and AWS Region. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Azure OpenAI | official | partial — Microsoft documents OpenAI client library usage with Azure endpoints, deployment names, and Azure-specific authentication; endpoint behavior and API versions can differ. | Usage-based pricing varies by model, deployment type, and Azure offer. Microsoft publishes Azure-specific prices; no single cross-provider markup is inferred. | Model and feature availability depends on deployment type and Azure region; consult Microsoft's model table. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Cerebras | official | partial — Cerebras documents OpenAI client support for common workflows and explicitly lists compatibility differences and unsupported parameter combinations. | Inference pricing and plan conditions are published by Cerebras and vary by model, tier, and availability. Platform-published inference pricing; not a uniform resale markup. | Only models in Cerebras’s current model catalog should be assumed available; exact identifiers and capabilities are model-specific. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Cloudflare Workers AI | official | partial — Cloudflare documents OpenAI-compatible endpoints for supported Workers AI models and API operations. | Usage-based pricing varies by model and usage unit, with documented allocation and billing rules. Cloudflare publishes platform pricing; no universal markup is inferred. | The supported catalog is defined by Cloudflare's model documentation and includes different task types. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Cohere | official | unknown — No current first-party OpenAI SDK compatibility statement was verified in this review; unknown does not mean incompatible. | Usage-based pricing varies by Cohere model and API workload; current rates and conditions are published on its pricing page. First-party published API pricing; not a resale markup. | Cohere’s official documentation describes its current model families and their intended API workloads; consult it for exact identifiers. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Databricks | official | yes — Databricks documents use of the OpenAI SDK with model-serving endpoints and Databricks authentication. | Pricing depends on the selected serving mode, model, and Databricks usage. Databricks publishes product-specific pricing; no universal markup is inferred. | Foundation Model APIs include supported pay-per-token and provisioned-throughput options; availability depends on workspace and region. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| DeepInfra | official | yes — DeepInfra documents an OpenAI-compatible chat completions endpoint using its base URL, API token, and catalog model names. | Pricing is specific to each model and billing unit; some tagged language models also have documented standard, priority, or flex service tiers. Platform-published model pricing; service tiers and workload units prevent a single universal comparison. | The public model catalog lists currently hosted models across multiple inference categories; consult each model page for capabilities and limits. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Featherless | official | yes — Featherless documents an OpenAI-compatible API with its own base URL and API key. | Subscription and usage terms depend on the plan documented by Featherless. Featherless publishes platform access pricing; no per-model markup is inferred. | The available serverless catalog is listed by Featherless and can change over time. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Fireworks AI | official | yes — Fireworks documents OpenAI Python and JavaScript SDK use with its base URL, including supported endpoints and request-field limitations. | Serverless inference uses model-specific token pricing; dedicated deployments and training use their documented workload-based pricing. Platform-published pricing varies by service and serving mode; no single markup applies. | The Fireworks model documentation covers supported text, vision, audio, image, and embedding models; availability is platform-specific. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| GitHub Models | aggregator | partial — GitHub documents OpenAI SDK usage for supported inference requests with GitHub authentication; not every OpenAI API feature is implied. | Usage is subject to GitHub Models rate limits and billing rules, which vary by model and usage tier. GitHub publishes usage and billing terms; no cross-provider markup percentage is inferred. | The GitHub Models catalog includes multiple model publishers; availability and limits are specific to GitHub Models. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Google AI Studio | official | partial — Google documents OpenAI library compatibility for the Gemini API with a changed base URL and configuration. | Model- and usage-dependent Gemini API pricing; free and paid usage tiers are documented separately. First-party pricing; tier and model conditions apply. | Available Gemini models and rate limits depend on the API and usage tier. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Google Vertex AI | official | partial — Google documents an OpenAI library compatibility path for supported Gemini calls on Vertex AI, with Google Cloud authentication and endpoint configuration. | Usage-based pricing depends on model, modality, and selected Vertex AI feature. Google Cloud publishes Vertex AI pricing; no cross-provider markup percentage is inferred. | Available models and features depend on the Vertex AI API, Google Cloud project, and region. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Groq | official | yes — Groq documents use of the OpenAI SDK with a Groq base URL and API key. | Usage-based pricing by supported model and token usage. Platform-published API pricing; not a uniform resale markup. | Only models listed in Groq’s current documentation should be assumed available. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Hugging Face Inference Providers | aggregator | partial — Hugging Face documents OpenAI client usage for supported chat completion requests; other APIs and provider-specific features are not implied. | Pricing depends on the selected model and inference provider, with provider-specific billing conditions. Provider-routed usage is not summarized as a single markup percentage. | Model availability and task support depend on the selected inference provider and the Inference Providers catalog. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Jina AI | official | unknown — General OpenAI API compatibility has not been verified from the official documentation for this profile. | Usage-based pricing varies by model, task, and billing unit. First-party published pricing; not a resale markup. | Available models and task-specific APIs are listed in Jina AI's documentation. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Lambda | official | yes — Lambda documents OpenAI-compatible usage for its Inference API with a Lambda endpoint and API key. | Inference pricing depends on the selected model and usage; compute products are priced separately. Lambda publishes product pricing; no universal markup is inferred. | The Inference API supports the models listed in Lambda's current documentation. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| MiniMax | official | unknown — General OpenAI API compatibility has not been verified from the official documentation for this profile. | Usage-based pricing varies by model and modality. First-party platform pricing; no resale markup is inferred. | Available models and modality-specific APIs are listed in MiniMax's platform documentation. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Mistral AI | official | unknown — No first-party OpenAI SDK compatibility statement was verified in this review; this does not mean the API is incompatible. | Usage-based API pricing varies by model and usage category; consult the current La Plateforme pricing table. First-party published API pricing; not a resale markup. | Mistral’s model overview documents available model families, capabilities, and API identifiers; availability can change. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Moonshot AI | official | yes — Moonshot documents use of the OpenAI SDK with its API base URL and API key for supported requests. | Usage-based pricing depends on the selected Moonshot model and billing item. First-party published API pricing; not a resale markup. | Available Kimi models and capabilities are listed in Moonshot's API documentation. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Nebius Token Factory | official | yes — Nebius documents OpenAI SDK usage with the Token Factory base URL and API key for supported models. | Model- and usage-dependent pricing is published for Token Factory. Platform-published model pricing; no universal resale markup is inferred. | The Token Factory model catalog defines currently supported hosted models and capabilities. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Novita AI | official | yes — Novita documents OpenAI-compatible API usage for supported language-model endpoints. | Usage-based pricing varies by model, modality, and API. Novita publishes platform prices; no uniform resale markup is inferred. | The supported model catalog and task coverage are listed in Novita's documentation. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| NVIDIA NIM | official | yes — NVIDIA documents OpenAI-compatible API usage for supported NIM language-model endpoints. | Access and pricing depend on the selected NVIDIA product, hosted API, or deployment option. No single markup applies across hosted APIs and self-deployed NIM products. | Supported models and capabilities are listed in NVIDIA's API catalog and individual NIM documentation. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Ollama | other | partial — Ollama documents OpenAI compatibility for supported local endpoints and operations, with compatibility limitations listed in its documentation. | Local model execution has no per-request Ollama API price; hardware, model licenses, and optional hosted products have separate terms. No hosted resale markup applies to local inference. | Models available for local use are listed in Ollama's library; actual availability depends on the user's downloaded models and hardware. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| OpenAI | official | yes — Native OpenAI API. | Usage-based, with separate input and output token prices by model. First-party published API pricing; not a resale markup. | The available catalog changes over time; consult the official model documentation. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| OpenRouter | aggregator | yes — OpenRouter documents an OpenAI-compatible API with its own base URL and optional features. | Model-specific pricing is exposed through the public models API and model pages. Not summarized as a single percentage; model and upstream pricing conditions differ. | The public models endpoint provides the platform catalog; availability is specific to OpenRouter. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Perplexity | official | yes — Perplexity documents use of the OpenAI SDK with its chat completions endpoint and API key. | Usage-based pricing depends on the selected model, tokens, and documented search-related charges. Perplexity publishes its API pricing; no resale markup is inferred. | Only models listed in Perplexity's current API documentation should be assumed available. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Replicate | official | unknown — General OpenAI API compatibility has not been verified from the official documentation for this profile. | Model-specific usage pricing is published on model pages; billing units vary by model. Replicate publishes per-model platform prices; no universal resale markup is inferred. | The platform hosts public and custom models with model-specific inputs, outputs, and billing units. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| SambaNova | official | yes — SambaNova documents an OpenAI-compatible API for supported chat and completion requests. | Pricing and access depend on the selected SambaNova product and usage plan. Platform-published access terms; no universal resale markup is inferred. | The current supported model list is maintained in SambaNova's documentation. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Together AI | official | yes — Together documents OpenAI-compatible API usage with a Together endpoint and API key. | Usage-based pricing varies by hosted model and workload. Platform-published pricing; not comparable to a uniform reseller markup. | The platform model catalog lists currently supported deployments. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Vercel AI Gateway | aggregator | yes — Vercel documents an OpenAI-compatible API and use of the OpenAI SDK with the AI Gateway base URL. | Model- and provider-specific usage pricing is published by the gateway. Gateway pricing and provider-specific conditions apply; no universal markup percentage is inferred. | The gateway model catalog lists supported providers and models; availability is specific to AI Gateway. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Voyage AI | official | unknown — General OpenAI API compatibility has not been verified from the official documentation for this profile. | Usage-based pricing depends on the selected embedding or reranking model and billing unit. First-party published pricing; not a resale markup. | Voyage AI documents its available embedding and reranking models; this is not a general chat-model catalog. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| xAI | official | partial — xAI documents OpenAI SDK integration with an xAI base URL and API key; xAI-specific features and API behavior still require its own documentation. | Usage-based API pricing depends on the selected Grok model and applicable token or capability charges. First-party published API pricing; not a resale markup. | The official xAI model documentation lists current Grok model identifiers, capabilities, and availability conditions. | Not verified | Not verified | 2026-10-24 UTC |
|---|
| Zhipu AI | official | unknown — General OpenAI API compatibility has not been verified from the official documentation for this profile. | Usage-based pricing varies by model and API on the BigModel open platform. First-party platform pricing; no resale markup is inferred. | Available GLM models and API capabilities are listed in the BigModel documentation. | Not verified | Not verified | 2026-10-24 UTC |
|---|