Quick StartModel List

Model Overview

OriginRouter aggregates dozens of models from OpenAI, Anthropic, Google, Zhipu, DeepSeek, MiniMax, Kimi, and more — covering conversation, reasoning, coding, embeddings, and image generation. This page gives you a quick overview of the model categories and how to choose between them.

Model Categories

International flagship chat models

Top-tier general conversation and reasoning, ideal for complex tasks, deep analysis, and high-quality content generation.

GPT-5.5 / GPT-5.4 / GPT-5.2 — OpenAI's latest generation, spanning flagship to efficient performance tiers
Claude Opus 4.6 / Sonnet 4.6 / Haiku 4.5 — Anthropic's Claude series; Opus for the most complex reasoning, Sonnet excels at coding, Haiku is fast and lightweight
Gemini 3.1 Pro / Flash — Google multimodal models with long context windows (up to 1M tokens)

Coding and reasoning models

Deeply optimized for code generation, math reasoning, and logical analysis.

GPT-5.4 / GPT-5.3 Codex — OpenAI's code-optimized series, great for understanding and refactoring large codebases
Claude Sonnet 4.6 — Top performer on coding benchmarks; the primary recommended model for the coding plan
DeepSeek V3.2 — Open-source model with strong competitiveness in coding and math reasoning

Domestic models

Unique advantages in Chinese comprehension, local knowledge, and cost-effectiveness.

GLM-5.1 / GLM-4.7 / GLM-4.5-Air — Zhipu's GLM series, strong Chinese understanding; Air is lightweight and efficient
Kimi K2.6 / K2.5 / K2 Thinking — Moonshot series; Thinking variant excels at deep reasoning
MiniMax M2.7 / M2.5 / M2 — MiniMax series with well-rounded general capabilities

Specialized models

Embedding models: text-embedding-3-large / text-embedding-3-small — convert text to vectors for semantic search and clustering
Image models: gpt-image-1 / gemini-2.5-flash-image-preview — generate images from text descriptions

How to Choose a Model

By use case

Use case	Recommended models	Notes
Everyday coding	`claude-sonnet-4-6`, `gpt-5.4`	Best coding experience, balanced performance and speed
Lightweight coding / completion	`claude-haiku-4-5`, `gpt-5.4-mini`	Fast responses, quota-efficient
Architecture design / complex reasoning	`claude-opus-4-6`, `gpt-5.5`	Maximum reasoning power
Chinese content creation	`GLM-5.1`, `kimi-k2.6`	Natural, fluent Chinese output
Long document processing	`gemini-3.1-pro-preview`	1M token context window
Image generation	`gpt-image-1`, `gemini-2.5-flash-image-preview`	Multimodal capability

Cost vs. performance trade-offs

Performance first: choose models on direct or hybrid routing, such as claude-opus-4-6 or gpt-5.5
Cost-efficiency first: use third-party channels or domestic models — costs can drop by about 85-95%
Quota-conscious: use claude-haiku-4-5 or GLM-4.5-Air for simple tasks, then switch to flagship models when needed

Routing strategies differ by subscription tier. Pro and above users get access to better routing channels and lower degradation probability. See Coding Plan and Model Routing Details.

Discover Available Models

Query via API

Call the GET /beta/v1/models endpoint to get the full list of currently available models, including model ID, owner, and creation time.

curl https://api.easytransnote.com/beta/v1/models \
  -H "Authorization: Bearer {YOUR_API_KEY}"

View API Reference

Browse in the console

Check the coding plan model table

Coding plan users can visit Model Routing Details to see the routing strategy for each model at different subscription tiers.

Model Routing

A single model on OriginRouter is usually connected to multiple upstream providers. You can use the system-recommended default route, or pin a specific provider by appending a route suffix. If the request does not specify a route suffix, the system automatically uses the model's default route to handle the call.

Default vs. Explicit

Default — when model is <model_id>, OriginRouter calls the model on its default provider.

{ "model": "claude-opus-4-6" }

Explicit — when model is <model_id>:<router>, the system uses that provider's instance of the model.

{ "model": "claude-opus-4-6:anthropic" }

Every model has a fixed list of supported providers. If you pass a router the model does not support, the system returns 404 invalid_model_origin. See Errors for details.

Supported Endpoints

Model routing is currently available on the following inference endpoints:

/beta/v1/chat/completions
/beta/v1/messages
/beta/v1/responses

If you call an endpoint that does not support routing, the system returns 404 invalid_model. See Errors for details.

Model Fallback

Beta endpoints enable experimental features by default, such as model fallback. Even with a valid <model_id>:<router>, your preferred provider may still fail (e.g. on internal thinking signatures or tool-call records) and the system will fall through to the next available provider.

You can force-disable fallback by setting fallback: "disabled" in the request body, or read more about model fallback.

To see which provider actually handled a request, check the response headers:

X-Originrouter-Model-Id — the model ID actually used
X-Originrouter-Actual-Route — the model provider actually used

Steps that fail during fallback are not directly billed.

Pricing

Different providers may price the same model differently. The Console → Model List page lists, for each model, the model ID, the default provider, all supported providers, and the price on each provider.

Next Steps

Now that you know what models are available, see how to access them at the best price through the coding plan.

Continue: Coding Plan