AI Model Services › Model Inference

Model Inference

AE-110.10 Level 2 API Evangelist extension API Evangelist extension, CC BY-NC-SA 125 providers 253 API surfaces 70 rated strong or better

Serving model calls on demand.

Model Inference (AE-110.10) is a level-2 business capability under AI Model Services in the API Evangelist extension. The catalog holds 253 API surface(s) from 125 provider(s) that can perform some part of it, 70 of them rated strong or better. Reach is the vendor surface that lands on this capability — it is not a claim about what any particular organisation has deployed.

Where this capability definition comes from. This capability is part of the API Evangelist extension to the business-capability model, written by API Evangelist and licensed CC BY-NC-SA 4.0. API Evangelist capability extension — https://apievangelist.com Audience: b2bagent It extends, and never modifies, Turbo EA Capabilities by Vincent Verdet (CC BY 4.0). · Extension provenance and licence

Sub-capabilities

Hosted LLM Inference AE-110.10.10API Evangelist extension, CC BY-NC-SA

b2bagent

Text generation behind an API.

no catalog coverage

OpenAI-Compatible Endpoints AE-110.10.20API Evangelist extension, CC BY-NC-SA

b2bagent

Inference behind the de facto chat-completions interface.

no catalog coverage

Model Routing & Gateways AE-110.10.30API Evangelist extension, CC BY-NC-SA

b2bagent

Choosing between models and providers per request.

no catalog coverage

On-Device & Edge Inference AE-110.10.40API Evangelist extension, CC BY-NC-SA

consumerb2b

Running models on phones, devices and the edge.

no catalog coverage

Providers that reach this capability

Ordered by rating band. Reach means a provider publishes an API surface that can perform some part of this capability — it is not a claim that any particular organisation has deployed it.

Exemplar 30 Complete, well-documented, and agent-ready
Strong 40 Solid coverage with minor gaps
Red Hat Ansible Automation PlatformAI1 API66.2FastlyChat Completions1 API66.0IBMMessagesText Generation2 APIs65.8PerplexityAsyncSonar2 APIs65.7GumloopChat Completions1 API65.6falStreaming1 API65.2Kimi (Moonshot AI)BatchChat CompletionsMessagesResponses4 APIs64.8Eden AIAnthropic MessagesChatDecisions3 APIs64.4CometChat Completions1 API63.6VercelChat Completions1 API63.1CoastyPredict1 API62.8AIMLAPIChat CompletionsMessagesResponsesThreads > Runs4 APIs62.6SAKURA internetChatMessagesResponses3 APIs62.6RightNow AIChat CompletionsResponses2 APIs62.5EdgeeChat CompletionsMessagesResponses3 APIs62.4brick.blueModels1 API62.2Cerebras SystemsChat CompletionsCompletions2 APIs62.1TaleModel endpointsThreads2 APIs62.0Dedalus LabsChat CompletionsResponses2 APIs61.9LangdockChat CompletionsFIMGoogleMessages5 APIs61.9Microsoft CopilotChat1 API61.3AppDirectAI ModelsChat CompletionsChat SessionsResponses API v24 APIs60.0H2O.aiChat1 API59.9NVIDIA NIMBiology1 API59.7TypeSafe AISystemone1 API59.7CONNSKILLEU-LLM1 API58.9Blazing Bullet Private LimitedCompletionsResponsesgoogle genai endpoints3 APIs58.8Sail ResearchModels API1 API58.6FriendliAIServerless.Model1 API58.1Venice.aiChat CompletionsResponses2 APIs58.0Altimate AIAgentsChat Completions2 APIs57.8PortkeyChat CompletionsCompletions2 APIs57.3AkkioInference1 API56.8InterfazeChat Completions1 API56.3SarvamChat Completions1 API56.2Arcade.devLLM1 API56.1gpt55-token-gatewayChat Completionsgpt55-model-gateway2 APIs56.0DiscountedTokens APIMessagesModelsResponses3 APIs55.2OpenRouterChat CompletionsCompletions2 APIs54.8AIHubMixAnthropic CompatibleGoogle Vertex AI Compatible2 APIs54.6
Developing 41 Usable, with meaningful gaps to close
Thin 12 Limited public surface area
Emerging 2 Early or largely undocumented
This page carries no rating. Capabilities are not rated. A capability is a description of what a business does, not a thing a company publishes, so a Kin Score would have nothing to measure.
The edge table is a Pro feature. This page shows which providers and tags reach Model Inference. The underlying tag → capability edges — each with the quoted fragment of the provider's own OpenAPI that evidences it, a calibrated confidence score, and the contract-provenance gate it passed — are available through the API, along with company-level capability maps. Only edges at confidence ≥ 0.7 with evidence found verbatim in the source contract are published at all.

See plans →  ·  How the edges are graded →