Model Inference APIs for AI Agents: Compare Providers, APIs, and Routes | AgentRouter

Model Inference APIs for AI Agents

Use AgentRouter to run chat completion, code completion, and model discovery across multiple providers through one routed model API.

DeepSeek

OOpenRouter

Anthropic

Groq

Live Providers

Live Routes

0 credits ($0)

Model Catalog

1 credit ($0.001)

Paid Inference

The models catalog currently exposes 3 capabilities for AI agents. Model Catalog currently starts at 0 credits ($0), and paid inference currently starts at 1 credit ($0.001).

Availability Note

The September 25, 2026 production sweep passed all 11 launched Models routes. Free OpenRouter discovery returned 459 models with estimated credits-per-million-token prices; DeepSeek returned `deepseek-flash` and `deepseek-v4-pro`; Groq chat returned `AgentRouter Groq smoke OK` from `openai/gpt-oss-20b`.

Supplier Pricing Snapshot

1,000 credits = $1 is the AgentRouter wallet conversion rate, not one OpenRouter model price. Each catalog row has its own estimated input/output credits per million tokens; every paid inference is settled per request from the final MPP quote returned as creditsCharged.

0 credits ($0)

OpenRouter Catalog

Free direct discovery with model ids, context, modalities, and public input/output price estimates.

From 1 credit · dynamic

OpenRouter Chat

OpenRouter is a gateway, not one model. The selected model and request size determine the final MPP quote.

4 credits ($0.004)

DeepSeek Chat

Latest paid smoke charged 4 credits on models.chat.complete.deepseek.mpp.

3 credits ($0.003)

DeepSeek Code/List

Latest paid smoke charged 3 credits on code completion and model listing.

API Reference

Three clear APIs: list, chat, and code

List All Models

Code samplelist-models.ts

import { AgentRouterClient } from "@agentrouter/agentrouter"; const client = new AgentRouterClient({  apiKey: process.env.AGENTIC_API_KEY,}); const catalog = await client.models.catalog.list(); console.log(catalog.modelCount);console.log(catalog.models[0]);

Sample response

{  modelCount: 458,  pricedModelCount: 452,  creditsCharged: 0,  models: [    {      id: "openai/gpt-5.6-sol",      contextLength: 1050000,      supported_parameters: ["max_tokens", "temperature", "top_p"],      agentRouterPricing: {        inputCreditsPerMillionTokens: 2000,        outputCreditsPerMillionTokens: 10000      }    },    "… live catalog continues"  ]}

Chat Completion

Code samplechat-complete.ts

import { AgentRouterClient } from "@agentrouter/agentrouter"; const client = new AgentRouterClient({  apiKey: process.env.AGENTIC_API_KEY,}); const result = await client.models.chat.complete.execute(  {    model: "anthropic/claude-sonnet-5",    messages: [      { role: "user", content: 'Reply only: "model route OK"' },    ],    max_tokens: 16,  },  {    routeKey: "models.chat.complete.openrouter.mpp",    allowFallback: false,  },); console.log(result);

Sample response

{  completionText: "model route OK",  model: "anthropic/claude-sonnet-5",  provider: "openrouter",  routeKey: "models.chat.complete.openrouter.mpp",  fallbackUsed: false,  creditsCharged: 12}

Code Completion

Code samplecode-complete.ts

import { AgentRouterClient } from "@agentrouter/agentrouter"; const client = new AgentRouterClient({  apiKey: process.env.AGENTIC_API_KEY,}); const result = await client.models.code.complete.execute(  {    model: "deepseek-flash",    prompt: "const label = ",    suffix: "\nconsole.log(label);",    max_tokens: 16,    temperature: 0,  },  {    routeKey: "models.code.complete.deepseek.mpp",    allowFallback: false,  },); console.log(result);

Sample response

{  completionText: "\"agentRouterSmoke\";",  model: "deepseek-flash",  provider: "deepseek",  routeKey: "models.code.complete.deepseek.mpp",  fallbackUsed: false,  finishReason: "stop",  creditsCharged: 3}

What Is AI Model Inference?

What is agent model inference for AI agents?

Agent model inference is the operational layer that lets an AI agent call raw language-model capabilities through a stable routed contract.

Direct model integrations tend to bind application logic to one provider's endpoint names, model ids, and billing mechanics. That makes it harder to compare suppliers or keep execution stable when upstream wrappers evolve.

AgentRouter keeps chat completion, code completion, and model listing behind one models domain so the agent can recommend, inspect route context, and execute without owning provider-specific payment details.

Run chat completion through a message-list contract
Complete code with prefix and optional suffix input
List live provider model identifiers before execution
Keep provider settlement behind one AgentRouter wallet

Top Scenarios

Top ai model inference scenarios for agents

Run a short model chat

Use chat completion when the workflow needs raw language-model output rather than search-grounded answers or hosted-agent execution.

Anthropic

MPP

1 credit ($0.001)

OpenRouter

MPP

1 credit ($0.001)

DeepSeek

MPP

4 credits ($0.004)

Model reasoning, transformations, and direct LLM responses

DeepSeek chat completion charged 4 credits in the May 5, 2026 production-mode smoke.

Coverage: 1 capabilities · 6 live routes

Start with Chat Complete

Complete code at the cursor

Use code completion when the agent has a prefix and optional suffix and wants a fill-in-the-middle result.

Code Complete

DeepSeek

3 credits ($0.003)

Code generation, FIM-style snippets, and editor workflows

DeepSeek code completion charged 3 credits in the May 5, 2026 production-mode smoke.

Coverage: 1 capabilities · 1 live routes

Start with Code Complete

Inspect available model ids

Use the free OpenRouter catalog before execution to compare live model ids, context windows, modalities, and public input/output token prices.

OpenRouter

Direct

0 credits ($0)

DeepSeek

MPP

3 credits ($0.003)

Groq

MPP

5 credits ($0.005)

Free model discovery, price comparison, and preflight checks

The broad OpenRouter catalog costs 0 credits. Provider-wrapper inventory routes remain available when a workflow needs that supplier's exact list.

Coverage: 1 capabilities · 4 live routes

Start with Models List

Provider Comparison

Compare ai model inference providers and APIs

The rows below are derived from the live route catalog currently exposed through AgentRouter.

Provider	Best For	Key Capabilities	Starting Price	Available Via
O OpenRouter Covers Chat Complete, Models List through the live routed catalog. 2 live routes	Chat Complete and Models List	Chat Complete, Models List Modes: Managed	0 credits ($0)	MPP, Direct
A Anthropic Covers Chat Complete through the live routed catalog. 2 live routes	Chat Complete	Chat Complete Modes: Managed	1 credit ($0.001)	MPP

Quick Start

How to integrate ai model inference with AgentRouter

Get one API key, list the live model ids and prices for free, then execute the exact model you want. AgentRouter selects a compatible provider route behind the same SDK call.

1
Step 1
Create one API key
Copy the key from Settings → API and expose it to the SDK as AGENTIC_API_KEY.
2
Step 2
List and choose a model
Call client.models.catalog.list() for live ids, context windows, and estimated input/output prices. Listing costs 0 credits.
3
Step 3
Execute that model
Pass the model id and messages to chat.complete.execute(). The response includes completionText and the final creditsCharged.

Checking AgentRouter account...

Works with

Pricing By Task

Pricing by common ai model inference tasks

These examples translate the current live route prices into workflow questions operators usually evaluate before shipping a domain into an agent.

Task	Route	Price	Notes
Run one chat completion	Anthropic (models.chat.complete.anthropic.messages.mpp)	1 credit ($0.001)	Use when the workflow wants raw model output from a message list.
Run one code completion	DeepSeek (models.code.complete.deepseek.mpp)	3 credits ($0.003)	Use when the workflow wants FIM-style completion between prefix and suffix.
List model ids	OpenRouter (models.list.openrouter)	0 credits ($0)	The default OpenRouter catalog route is free and includes public prices converted to AgentRouter credits per million tokens.

Pricing note: 1000 credits = $1 USD. Raw API responses still return numeric credit fields such as creditsCharged.

Why AgentRouter

Why use AgentRouter instead of direct ai model inference integrations?

Choose an exact model when you already know what you want. If you only know the task, the optional recommend call can rank compatible routes by cost and reliability. Either way, the same execute contract stays stable while AgentRouter handles provider routing and settlement.

One key to list models, inspect prices, and execute the selected model.
Direct model choice for the common case; optional recommendation when the model is unknown.
One wallet and one final creditsCharged field across routed providers.
Full route contracts and provider details stay in the reference section below.

Jump to detailed capabilities

FAQ

AI Model Inference FAQ

What belongs in the models domain?

Raw language-model inference belongs here: chat completion, code completion, and model discovery. Search-grounded answers, hosted agents, speech, translation, images, and document understanding stay in their dedicated domains.

Do I need a direct DeepSeek API key?

No. The live DeepSeek routes use AgentRouter's MPP settlement path and execute through one AgentRouter wallet.

Which model ids are currently exposed?

Call the free models.list OpenRouter route or client.models.catalog.list() for the current catalog instead of relying on a hard-coded list. Provider-specific routes still document accepted aliases and exact wrapper inventory in route context.

Does the user need to fund Tempo or hold provider API keys?

No. The user spends AgentRouter wallet credits. AgentRouter owns provider routing and backend settlement; the operator maintains the Base and Tempo settlement balances shown in the admin dashboard.

Reference

Detailed ai model inference capabilities and route pricing

Use the capability browser below when you want contract-level detail: endpoints, live example modals, route breakdowns, and exact route prices.

Provider Networks

Direct API Partner·

AgentRouter·x402·

Capability	Endpoints	Price
chat.complete Complete a chat conversation with an OpenAI-compatible message list.	`GET /api/agentic-api/domains/models/capabilities/chat-complete` Read the capability contract, current lifecycle status, and machine-readable metadata for this model inference workflow.	Free
	`POST /api/agentic-api/domains/models/capabilities/chat-complete/recommend` Ask AgentRouter to compare eligible routes inside this capability and return the best recommendation before execution.	Free
	`POST /api/agentic-api/domains/models/capabilities/chat-complete/execute` Execute this capability through the selected route while AgentRouter handles pricing, wallet debits, and upstream settlement.
code.complete Complete code with fill-in-the-middle style prefix and optional suffix input.	`GET /api/agentic-api/domains/models/capabilities/code-complete` Read the capability contract, current lifecycle status, and machine-readable metadata for this model inference workflow.	Free
	`POST /api/agentic-api/domains/models/capabilities/code-complete/recommend` Ask AgentRouter to compare eligible routes inside this capability and return the best recommendation before execution.	Free
	`POST /api/agentic-api/domains/models/capabilities/code-complete/execute` Execute this capability through the selected route while AgentRouter handles pricing, wallet debits, and upstream settlement.
list List model identifiers and availability metadata exposed by the provider wrapper.	`GET /api/agentic-api/domains/models/capabilities/list` Read the capability contract, current lifecycle status, and machine-readable metadata for this model inference workflow.	Free
	`POST /api/agentic-api/domains/models/capabilities/list/recommend` Ask AgentRouter to compare eligible routes inside this capability and return the best recommendation before execution.	Free
	`POST /api/agentic-api/domains/models/capabilities/list/execute` Execute this capability through the selected route while AgentRouter handles pricing, wallet debits, and upstream settlement.

Related AgentRouter pages

AgentRouter landing page

Browse all live domains, top API cards, and route coverage across the platform.

Install and auth

Enable AgentRouter, create the API key, and connect the shared wallet flow.

Phone, SMS & Voice domain

Create managed phone agents, provision numbers, inspect conversations and calls, and keep telephony workflows behind one supplier-neutral phone domain.

Physical Mail domain

Quote letters, send physical mail, poll fulfillment status, and send postcards through one routed post domain.