Model Guide

What Is a Model

A model is the core engine that powers an AI member's ability to think, understand, and generate content. Different models are trained by different AI companies and vary in speed, reasoning ability, language understanding, and cost. Choosing the right model helps you balance output quality and usage cost.

StoryClaw integrates large language models (LLMs) from major AI providers including Anthropic, DeepSeek, Google, OpenAI, xAI, Moonshot, and MiniMax. It also provides three intelligent routing modes (auto / eco / premium) that automatically route your request to a suitable underlying model based on task type — you are charged at that model's actual rate.

How to Switch Models

At the bottom-left of the chat input area in ClawBot, you can see the name of the currently active model (highlighted in red below). Click it to open the model selection panel.

Clicking opens the model selector with a search field. The right side shows each model's input/output price per 1M tokens.

Intelligent Routing Modes

StoryClaw offers three routing modes prefixed with storyclaw/. The platform automatically selects a suitable underlying model based on task type — no manual selection needed. The actual cost is determined by whichever underlying model the system routes to; there is no additional surcharge for the routing itself.

storyclaw/auto

• Billing: Billed at the routed model's rate
• Strategy: Platform automatically routes to the most suitable underlying model based on task type
• Best For: General use; default when unsure what to pick

storyclaw/eco

• Billing: Billed at the routed model's rate
• Strategy: Economy mode — routes to faster, lower-cost underlying models
• Best For: Simple Q&A, high-frequency or batch tasks, cost-sensitive scenarios

storyclaw/premium

• Billing: Billed at the routed model's rate
• Strategy: Premium mode — routes to more powerful flagship models
• Best For: Complex creative tasks, in-depth analysis, quality-critical scenarios

Understanding the Price Format

The price displayed next to each model follows the format: Input price / Output price /1M — the cost per one million tokens processed. For example, $3 / $15 /1M means: $3 per 1M tokens you send to the model (input), and $15 per 1M tokens the model generates (output).

Model Details

If you want precise control over which model is used, you can select one directly in the model selection panel. The main LLM models currently supported are listed below, grouped by provider. The actual available models are subject to what's shown in the console — the list is updated continuously as new models are added.

Anthropic — Claude Series

storyclaw/claude-opus-4-6

Flagship · $5 / $25 /1M · 1M ctx

Best For: Complex analytical reports, reviewing long contracts or legal documents, designing multi-step agent workflows, high-quality brand copywriting

Anthropic's flagship model with rigorous logical reasoning, precise language output, and excellent performance on complex instructions and long-form text.

storyclaw/claude-opus-4-7

Flagship · $5 / $25 /1M · 1M ctx

Best For: Step-by-step logic and math problems, complex strategic planning and decision analysis, multi-agent collaboration tasks, highest-quality creative writing

Latest Claude Opus with Extended Thinking — the model reasons step-by-step before responding, delivering notably higher accuracy on multi-step reasoning tasks.

storyclaw/claude-sonnet-4-6

Balanced · $3 / $15 /1M · 1M ctx

Best For: Everyday content creation and polishing, conversational assistants, meeting notes and document summaries, moderately complex analysis

The most balanced model in the Claude family — noticeably faster than Opus while maintaining first-tier output quality; best value for everyday use.

DeepSeek — DeepSeek Series

storyclaw/deepseek-v4-flash

Fast · $0.14 / $0.28 /1M · 1M ctx

Best For: Automated batch processing, simple Q&A chatbots, high-frequency workflow trigger nodes, highly cost-sensitive workloads

DeepSeek's lightweight fast model with low latency and among the lowest token costs available — ideal for high-concurrency scenarios with strict speed and cost requirements.

storyclaw/deepseek-v4-pro

Reasoning · $0.435 / $0.87 /1M · 1M ctx

Best For: Writing and debugging code (Python, JavaScript, etc.), solving math or algorithm problems, data processing and structured analysis, strong-reasoning tasks on a budget

DeepSeek's flagship reasoning model — excels at code generation, mathematical reasoning, and logical analysis; strong overall capability at a highly competitive price.

Google — Gemini Series

storyclaw/gemini-3-flash-preview

Balanced · $0.5 / $3 /1M · 1M ctx

Best For: Quickly summarizing large document sets, article generation and polishing, general-purpose Q&A, scenarios requiring fast processing of very long text

Google's latest fast language model with balanced speed and capability, supporting up to 1M token context for large-scale text processing in a single request.

storyclaw/gemini-3.1-flash-lite-preview

Lite · $0.25 / $1.5 /1M · 1M ctx

Best For: Text classification and key information extraction, simple translation and format conversion, high-concurrency low-cost pipelines, preliminary content screening

Lightweight variant of Gemini Flash that significantly reduces cost while retaining basic language capabilities — suited for tasks where top-tier quality isn't required.

OpenAI — GPT Series

storyclaw/gpt-5.4

Balanced · $2.5 / $15 /1M · 1M ctx

Best For: Agent tasks integrating external APIs, multi-tool collaborative workflows, reliable content generation and editing, business automation requiring tool-use capabilities

A balanced GPT-5 generation model with accurate instruction understanding, stable overall performance, and solid tool-use support — well-suited for agent scenarios with external system integrations.

storyclaw/gpt-5.5

Flagship · $5 / $30 /1M · 1M ctx

Best For: High-quality creative writing, complex business strategy analysis and proposal writing, demanding multi-step reasoning, quality-critical scenarios with sufficient budget

OpenAI's flagship model with top-tier performance across reasoning, creative writing, and instruction understanding — ideal for demanding tasks where output quality is paramount.

xAI — Grok Series

storyclaw/grok-4.20

Reasoning · $2.5 / $5 /1M · 1M ctx

Best For: In-depth research reports, synthesizing information from multiple sources, complex analysis requiring critical thinking, dialectical reasoning and comparative argument tasks

xAI's flagship reasoning model with deep thinking capability — excels at multi-perspective analysis and critical reasoning on complex problems, at a more competitive price than most flagship models.

Moonshot — Kimi Series

storyclaw/kimi-k2.5

Economy · $0.6 / $3 /1M · 256K ctx

Best For: Chinese article writing and editing, reading comprehension and summarization of uploaded documents, enterprise knowledge base Q&A, cost-sensitive Chinese content workflows

A high-value model from Moonshot with strong Chinese language understanding and generation, performing reliably on document reading and knowledge Q&A.

storyclaw/kimi-k2.6

Balanced · $0.95 / $4 /1M · 256K ctx

Best For: Long-form Chinese reports and proposals, intelligent enterprise knowledge base retrieval and Q&A, Chinese AI workflows requiring precise instruction-following, multi-step Chinese agent tasks

Enhanced Kimi with improved Chinese reasoning and instruction-following, plus better support for multi-step agent collaboration — suited for Chinese workflows requiring precise complex instruction understanding.

MiniMax — MiniMax Series

storyclaw/MiniMax-M2.7

Economy · $0.3 / $1.2 /1M · 200K ctx

Best For: Bulk social media content in Chinese, mass product description writing for e-commerce, budget-constrained Chinese chatbots, cost-first Chinese content automation

A domestically developed LLM with strong Chinese text generation and among the lowest token costs — designed for price-sensitive, high-volume Chinese content production.

Quick Selection Guide

Not sure which model to pick? Match your task type to the table below for a quick decision: