LLM Capability Leaderboard

OpenAI frontier work model for complex real-world tasks, agentic coding, research, data analysis, and cross-tool execution.

General model with strong long-context, multimodal, and reasoning capability.

Frontier general model for complex reasoning, coding, tool use, and professional writing.

Premium model with strong writing, code review, and agentic workflow behavior.

Z.AI flagship model for long-horizon agent tasks, real-world engineering delivery, coding, and complex reasoning.

DeepSeek next-generation V4 family model for long-context work, coding, math reasoning, and cost-efficient production APIs.

Z.AI GLM-5 foundation model for general reasoning, writing, coding, and agentic engineering workflows.

Coding-optimized model for repository-scale editing, testing, and debugging.

Thinking model for complex Q&A and fresh-information workflows.

Low-latency DeepSeek V4 variant for high-throughput, cost-sensitive, and real-time product scenarios.

Balanced multilingual and coding model with broad API availability.

Open-weight reasoning model focused on math, coding, and cost-efficient deployment.

Z.AI GLM-4.6 improves real-world coding, long-context processing, reasoning, search, writing, and agentic applications.

General model with strong long-context and Chinese-language performance.

European model for enterprise API, coding, and multilingual workloads.

Large open-weight model for self-hosted enterprise and research workflows.

Stable model for Chinese, writing, and general knowledge tasks.

Open-weight MoE model for self-hosting and cost-sensitive inference.