Overview

The Turing Platform is TCL's enterprise-grade AI infrastructure. A single API Key gives you access to large language models, multimodal generation, and search capabilities from dozens of providers—including OpenAI, Anthropic, Google, Alibaba, ByteDance, Zhipu, DeepSeek, KIMI, and MiniMax. The platform handles:

• Unified access layer: A single OpenAI-compatible /chat/completions endpoint reaches every provider—just swap the model field. Claude also supports the native /messages endpoint.
• Elastic routing and fallback: Automatically switches to a backup model when the primary is unavailable; routes by quota and region.
• Billing and usage tracking: Every call is traced; billing is precise down to tokens, cache hits, and search queries.
• Security and compliance: Independent nodes for the China region, Europe region, and North America region—logs and model inference never leave the designated domain.

Platform Capabilities at a Glance

Capability	Endpoint	Supported Providers
Chat / Reasoning	POST /v1/chat/completions · POST /v1/messages · POST /v1/responses	Per the API fields of each model in the Model Catalog
Reasoning Models	Same as above + reasoning_effort / thinking parameters	GPT-5 · Claude 4.6 / 4.7 · DeepSeek R1 · Qwen Thinking Edition
Web Search	tools field in /chat/completions	Qwen enable_search · Gemini googleSearch · Claude web_search_20250305
Prompt Caching	Caller-side cache_control or automatic caching	Claude explicit cache_control; OpenAI GPT-4o/5 automatic
Image Generation	POST /v1/images/generations · POST /v1/chat/completions (Gemini)	Seedream · Gemini Nano Banana · Seedance
Video Generation	Dedicated endpoint	Veo · Seedance 2.0
Embeddings	POST /v1/embeddings	OpenAI · Alibaba · Zhipu · and more
Reranking	Dedicated endpoint	Cohere Rerank · Aliyun Rerank
Speech	POST /v1/audio/speech · POST /v1/audio/transcriptions/runs · Realtime API	OpenAI · Azure · Alibaba Cloud Tingwu

Quick Index

Getting started

• First-time integration → Quick Start
• Choosing a model → Model List
• Making your first call → First Call

Find docs by capability

• Chat endpoint usage → Chat Completions
• GPT-5 reasoning / MCP → OpenAI Response API
• Claude-native features (caching / reasoning / web_search) → Anthropic Messages
• Reasoning mode configuration → Reasoning Models
• Web search → Model Built-in Web Search
• Prompt caching → Claude Prompt Caching

Operations

• Endpoint schema + Try-It → API Reference
• Billing / usage / reconciliation → Billing and Usage
• Timeouts / retries / fallback → Timeouts, Retries, and Fallback
• Request tracing and troubleshooting → Request Tracing

Enterprise onboarding

• Industry account application → Industry Integration
• AI coding tools → Claude Code / Codex / OpenCode and more

Protocols and SDKs

The platform supports two compatible protocols. The protocols actually supported by each model are indicated by the API fields in the Model Catalog:

Protocol	Endpoint	SDK	Best for
OpenAI-compatible	/v1/chat/completions · /v1/responses · /v1/embeddings · /v1/images/generations	Official or community OpenAI SDKs (Python / Java / Node.js / Go / Ruby / ...)	Cross-provider unification; existing OpenAI ecosystem code
Anthropic native	/v1/messages	Official Anthropic SDKs (Python / Node.js / Java)	Claude-exclusive features such as cache_control, native thinking blocks, and web_search_20250305

SDKs can be installed directly—the Turing Platform does not pin SDK versions. Simply point base_url at the Turing node and replace api_key with your platform key. Installation instructions and first-call examples for each language are available in Quick Start → First Call (cURL / Python / Java / JavaScript samples with one-click copy).

For other languages (Go / Ruby / C# / PHP / ...), use the respective OpenAI or Anthropic community SDK for that ecosystem, or call the API directly with cURL. The API Reference provides Try-It functionality and code samples for these languages on each endpoint page.