Model families

Leading international providers

OpenAI (GPT and o series)

OpenAI's GPT family established the industry pattern of pre-training and conversational interfaces: GPT-3 popularized prompting, GPT-4 expanded multimodal use, and the current generation is GPT-5.x. gpt-5.6-sol is a general-purpose flagship with strengths in coding and multi-agent collaboration; gpt-5.6-terra targets specialized coding. Reasoning tasks on the platform can use claude-opus-4-8, gpt-5.5, deepseek-v4-pro and similar models (the industry o series is not currently offered).

Anthropic(Claude)

Anthropic is known for safety-alignment research through Constitutional AI. Claude excels at long-document analysis, writing and agent tasks. In the current generation, claude-opus-4-8 targets top-tier reasoning and complex agent tasks, while claude-sonnet-5 balances speed and intelligence as a common production-agent choice, with capabilities such as Extended Thinking.

Google(Gemini)

Google DeepMind's Gemini family is known for native multimodality and very long context, with direct video and audio understanding (an industry description). The platform currently offers Gemini 3.x models such as gemini-3.7-flash, gemini-3.5-flash, gemini-3.6-flash and gemini-3.1-pro; see the Model Catalog for availability.

Meta(Llama)

Meta's Llama family is a foundation of the open-source ecosystem, supporting the growth of deployment tools such as Ollama and vLLM and the fine-tuning ecosystem (an industry description; the platform does not currently offer Llama models).

Mistral

Mistral is a leading European provider known for efficient architectures, open weights and balanced multilingual capabilities (an industry description; the platform does not currently offer Mistral models).

xAI(Grok)

xAI's Grok family is closely integrated with X, emphasizing realtime information and a distinctive conversational style, with deep-search capabilities. Grok 4 represents its current generation.

Leading Chinese providers

DeepSeek

DeepSeek gained worldwide attention for open-source MoE architectures and strong value. Its R-series reasoning models helped increase the industry's investment in deep reasoning. The current generation is DeepSeek V4: deepseek-v4-pro is an open-source MoE flagship with strong coding capabilities; deepseek-v4-flash targets lightweight, low-cost uses.

Alibaba (Qwen)

Qwen offers one of China's most comprehensive model lineups, combining open-source models and flagship APIs with strong multilingual and tool-calling capabilities (an industry description; the platform does not currently offer Qwen models).

Zhipu (GLM)

Zhipu originated from a Tsinghua University technical team. Its GLM family combines strong Chinese understanding, agents and tool calling with an open-source tradition. GLM-5 is the current representative family.

Baidu (ERNIE)

Baidu's ERNIE was among the earliest conversational large-model products in China. It emphasizes knowledge enhancement and integration with Baidu's ecosystem. Some versions have also become open source in recent years; ERNIE is its representative model family.

Selection recommendations

Quick use-case guide:

• General use and coding: gpt-5.6-sol, claude-sonnet-5, deepseek-v4-pro
• Deep reasoning: claude-opus-4-8, gpt-5.5
• Very long context and multimodality: kimi-k3, gpt-5.6-sol
• Dedicated coding: gpt-5.6-terra, deepseek-v4-pro
• Cost-sensitive tasks: deepseek-v4-flash, glm-5.2

See the Model Catalog for the models currently offered and their prices.