​
Text Model Support Matrix

The complexity of the text series lies not in “having many models,” but in “having many protocols.” The current system supports OpenAI Chat Completions, OpenAI Responses, Claude Messages, Gemini native format, Legacy Completions, and Realtime at the same time. Without a matrix, external users can easily see only the API name, without knowing which protocol they should integrate with.
This page is organized by “protocol style + capability type,” rather than listing models one by one. Which models are actually available to a single API key is still determined by the result returned from Model List.

​
Protocol Overview

Protocol/CapabilityRecommended EntryTypical Model FamiliesStreamingTool CallingMultimodalUse Cases
OpenAI Chat Completions/v1/chat/completionsGPT, Claude mapped models, Gemini mapped models, DeepSeek, Qwen, etc.SupportedSupportedSupportedChat, Agent, legacy OpenAI SDK compatibility
OpenAI Responses/v1/responsesGPT, compatibility models that support Responses semanticsSupportedSupportedSupportedStructured output, tool orchestration, context continuation
Claude Messages/v1/messagesClaude Sonnet / Opus / Haiku, etc.SupportedSupportedSupportedExisting Anthropic SDK or Claude native prompt structure
Gemini Native/v1beta/models/{model}:{action}Gemini Flash / Pro / image preview modelsSupportedSupportedSupportedExisting Gemini SDK, preserving contents/parts structure
Legacy Completions/v1/completionsText models compatible with the legacy Completion styleUsually not preferredLimitedNoLegacy system compatibility
Realtime/v1/realtimeReal-time conversational modelsReal-time streamingDepends on modelMultimodal depends on the modelWebSocket real-time interaction

​
Common Model Name Examples

The following are not a “complete list of available models,” but examples of common text model names that are most likely to be encountered externally in the current system. Actual availability still depends on your channel configuration and the result returned by GET /v1/models.

​
OpenAI / GPT Style

Common Model NameDescription
gpt-4oGeneral-purpose multimodal model
gpt-4o-miniLightweight general-purpose model
gpt-4.1General-purpose text/tool-calling model
gpt-4.1-miniLightweight 4.1
gpt-4.1-nanoSmaller 4.1 variant
o1Reasoning model
o3Reasoning model
o4Reasoning model

​
Claude

Common Model NameDescription
claude-sonnet-4-20250514Common version in the Claude Sonnet family
claude-sonnet-4-5-20250929Later Sonnet version
claude-sonnet-4-6Sonnet 4.6
claude-opus-4-20250514Claude Opus family
claude-opus-4-1-20250805Later Opus version
claude-opus-4-6Opus 4.6
claude-opus-4-7Opus 4.7
claude-haiku-4-5-20251001Lightweight Haiku model

​
Gemini

Common Model NameDescription
gemini-2.0-flashGemini general fast model
gemini-2.5-flashGemini 2.5 Flash
gemini-2.5-proGemini 2.5 Pro
gemini-3.1-pro-previewGemini preview Pro model
gemini-3-pro-image-previewGemini image preview model
gemini-3.1-flash-image-previewGemini image preview fast model

​
DeepSeek

Common Model NameDescription
deepseek-chatGeneral chat model
deepseek-reasonerReasoning model
deepseek-v4-flashNew-generation fast model
deepseek-v4-proNew-generation advanced model
deepseek-r1Reasoning series
deepseek-r1-distill-qwen-32bDistilled reasoning model

​
Qwen

Common Model NameDescription
qwen-maxHigh-spec Qwen model
qwen3-maxHigh-spec model in the Qwen 3 family
qwen3-omni-flashQwen multimodal fast model
qwq / qwq-*Common naming for the Qwen reasoning series

​
Choose an Entry Point by Protocol

​
1. OpenAI Chat Completions

ItemDescription
EntryPOST /v1/chat/completions
Typical Fieldsmodel, messages, stream, tools, response_format
Multimodal Inputmessages[].content supports text, image_url, input_audio, file, video_url
Best ForUsers already using the OpenAI SDK and wanting the lowest migration cost
Typical Model FamiliesGPT, Claude mapped models, Gemini mapped models, DeepSeek, Qwen
Common Model Name Examplesgpt-4o, gpt-4.1, gpt-4o-mini, deepseek-chat, deepseek-reasoner, qwen-max

​
2. OpenAI Responses

ItemDescription
EntryPOST /v1/responses
Typical Fieldsmodel, input, instructions, tools, previous_response_id, text
AdvantagesBetter suited for structured output, tool calling, and complex task orchestration
StreamingSupports SSE
Typical Model FamiliesGPT-style models and upstream models compatible with Responses semantics
Common Model Name Examplesgpt-4o, gpt-4.1, o1, o3

​
3. Claude Messages

ItemDescription
EntryPOST /v1/messages
Typical Fieldsmodel, messages, system, max_tokens, tools, thinking
Auth CompatibilityAuthorization: Bearer or x-api-key + anthropic-version
Best ForUsers who already have Claude native-format calling code
Common Model Name Examplesclaude-sonnet-4-20250514, claude-sonnet-4-6, claude-opus-4-6, claude-opus-4-7

​
4. Gemini Native

ItemDescription
EntryPOST /v1beta/models/{model}:generateContent
Typical Fieldscontents, generationConfig, tools, safetySettings
Auth CompatibilityAuthorization: Bearer or x-goog-api-key
Best ForCases where you need to fully preserve the Gemini contents/parts structure, image parts, and Google tool style
Common Model Name Examplesgemini-2.0-flash, gemini-2.5-flash, gemini-2.5-pro, gemini-3.1-pro-preview

​
Choose an Interface by Capability

​
Conversation Generation

NeedRecommended InterfaceDescription
General chat/v1/chat/completionsThe most general-purpose and compatible
Tool orchestration/v1/responsesBetter suited for structured workflows
Claude native/v1/messagesPreserve the Anthropic request structure
Gemini native/v1beta/models/{model}:generateContentPreserve the Gemini request structure

​
Realtime and Model Query

NeedRecommended InterfaceDescription
Real-time connection/v1/realtimeWebSocket real-time session
Query available models for the current key/v1/modelsReturns the default OpenAI-style response
Query Gemini native models/v1beta/modelsPreserves the Gemini native model format

​
Common Model Family Mapping Approach

FamilyCommon IntegrationDescription
GPT / OpenAI-style models/v1/chat/completions or /v1/responsesBy default, prioritize OpenAI-compatible protocols
Claude/v1/messages or /v1/chat/completionsFor new integrations, messages is recommended; the compatibility layer can also use chat
Gemini/v1beta/models/{model}:{action} or /v1/chat/completionsUse v1beta for native capabilities, or chat for unified SDK compatibility
DeepSeek / Qwen and other OpenAI-style upstream models/v1/chat/completionsUsually the simplest option

​
Selection Recommendations

  1. If you are already an OpenAI SDK client, start with the General Chat Interface (default streaming).
  2. If you need tool calling, context continuation, or structured output, prefer the OpenAI Multimodal Responses Interface.
  3. If you already have the Anthropic SDK, connect directly to the Claude Messages Interface.
  4. If you already have the Gemini SDK or need to preserve the contents/parts structure, connect directly to the Gemini Native Format.
  5. If you need low-latency bidirectional voice or a real-time session, check the Realtime Interface first.
  6. If you are not sure which models your current key can use, check the Model List first.