模型 API公开报价

原站公开的站内额度标价；充值折扣、规格附加费及最终扣费以原站为准。

原站公开的站内额度标价；充值折扣、规格附加费及最终扣费以原站为准。

原站公开的站内额度标价；充值折扣、规格附加费及最终扣费以原站为准。

原站公开的站内额度标价；充值折扣、规格附加费及最终扣费以原站为准。

原站公开的站内额度标价；充值折扣、规格附加费及最终扣费以原站为准。

原站公开的站内额度标价；充值折扣、规格附加费及最终扣费以原站为准。

原站公开的站内额度标价；充值折扣、规格附加费及最终扣费以原站为准。

原站按任务参数或阶梯规则计价，请在原站选择具体规格。

A pinned Sonnet 4.5 snapshot that balances coding, analysis, agent work, and spend. The dependable default when you need both quality and restraint. Undated Sonnet 4.5 name for coding and agent workflows.

The frontier GPT to start with for coding, computer use, research, and knowledge work. Broadly capable enough to be the default without feeling bland.

The frontier GPT to start with for coding, computer use, research, and knowledge work. Broadly capable enough to be the default without feeling bland.

A balanced Claude for code, analysis, and agent workflows, with an eye on the bill. A sensible everyday choice before reaching for Opus.

原站按任务参数或阶梯规则计价，请在原站选择具体规格。

A pinned Haiku 4.5 snapshot for fast assistance, classification, and compact agent jobs. Small, responsive, and easy to scale.

The Claude workhorse for coding agents, careful analysis, and keeping production costs sane. Not flashy; reliably useful.

GLM-5.1 is a text model from Zhipu for sustained tasks, reasoning, long documents, code and tool use. It offers about 200K context and up to 128K output, with a separate price tier above 32K input tokens.

原站公开的站内额度标价；充值折扣、规格附加费及最终扣费以原站为准。

Top-shelf Opus for the hardest reasoning, coding, and long-running agents. Reserve it for work that deserves the extra thought.

Claude Opus 5.5 is designed for sustained agentic coding and knowledge work. It supports text/image input, 1M context, up to 128K output and always-on adaptive thinking.

超复科技自研法律专家

A stronger Opus tier for serious software work and decisions where shallow answers are expensive. Bring it the messy, high-stakes problems.

Claude Sonnet 5.5 balances response speed and reasoning for coding, tool use and multi-step agents. It supports text/image input, 1M context, up to 128K output and adaptive thinking.

Flagship Claude for deep reasoning, demanding code, and agents with a long memory of the job. The heavyweight when the task is not routine.

原站公开的站内额度标价；充值折扣、规格附加费及最终扣费以原站为准。