Sea Whale AI

What is Sea Whale AI?

SeaWhale AI is a multi-model AI service designed for both ordinary users and developers.

Regular users can use functions for conversation, writing, academic tasks, and drawing, while developers can access the model through a unified API.

Two product entry points

The membership levels, computing power, account balances, and pricing for the two sets of services cannot be combined into a single set of quotas.

Main functions

• Access to models such as Tongyi Qianwen, DouBao, WenXin, Zhipu, and DeepSeek.
• Conduct multiple rounds of dialogue, translation, and content creation.
• Use AI writing templates for different scenarios.
• Upload documents and extract, locate, and summarize their content.
• Complete the full translation, article polishing, and academic review.
• Generate a paper outline and a draft for reference.
• Create materials using text-to-image, image-to-image, and image editing tools.
• Use assistants, Skills, and MCP to carry out specific tasks.
• Browse tutorials and public content in knowledge communities.
• Access hundreds of models through a unified API.

Guides to Using Haijing AI

1. Open the regular user dashboard and complete login by scanning a code or entering a phone number.
2. Choose dialogue, writing, academic, or drawing based on the task.
3. Compare capabilities, speed, and computational resource consumption in the model list.
4. Specify the task background, objectives, and output format.
5. Remove passwords and sensitive personal information before uploading files.
6. First, verify the effectiveness using short texts or small files.
7. Check the precomputed capacity before submitting long tasks.
8. Manual verification of facts, citations, data, and generated images is carried out.
9. Save the desired results and remove any data that is no longer in use.
10. To develop integration, one must go to the API platform to create a key separately.

Multi-model dialogue

SeaWhale AI aggregates multiple models, allowing users to choose based on Chinese language, reasoning, code, long texts, or cost.

The names and versions of third-party models will be updated; the fact that they are aggregated on this platform does not mean that these models were developed by Haijing AI.

AI writing

The writing module covers scenarios such as papers, outlines, work summaries, plans, marketing strategies, and text optimization.

• First, clarify the audience, channels, and purposes.
• Provide the true information that must be retained.
• It is required to first output the outline before writing the main text.
• Long texts are generated chapter by chapter with consistent terminology.
• Remove fictional cases, data, and sources.
• The final version is manually revised by the author, who assumes full responsibility for it.

AI academic assistant

The academic module offers document analysis, full-text translation, polishing, and assistance with peer review.

The public page lists the different formats in which documents can be presented, such as PDF, DOC, plain text, and links.

Tutorial for document parsing

1. Confirm that you have the right to upload and process documents.
2. Select AI document parsing and upload the file or link.
3. First, let the system generate the directory and content summary.
4. Ask questions regarding specific chapters, data, or tables.
5. Indicate the corresponding position in the original text for the required answer.
6. Check the original document for formulas, conclusions, and references.
7. A fixed glossary should be created when translation is required.
8. Remove the unsubstantiated content added to the model before exporting the results.

AI paper writing tools

The paper module can generate an outline from the title and educational level, which can then be expanded to serve as a reference for the initial draft of the paper.

• Supports selection at the levels of specialist, bachelor’s, master’s, and doctoral degrees.
• You can choose between a secondary or tertiary outline.
• Covers a variety of specialties and different lengths.
• Templates for summaries, literature reviews, and acknowledgments can be generated.
• A 1,000-word outline can be generated for free upon first use.
• The generated sample papers should not be submitted or published directly.

Price of papers

The following are the reference details made public on the official paper page as of August 31, 2026.

The price, word count, additional benefits, and final amount are based on the order page prior to generation.

Academic integrity

AI can assist in organizing structure and language, but it cannot replace genuine research, experiments, and the author’s contributions.

• Comply with the policies of schools and journals regarding AI tools.
• Do not fabricate data, interviews, experiments, or references.
• Verify each reference to ensure it indeed exists.
• Cited content should be marked with proper notation.
• Retain the outline, original data, and modification records.
• Disclose the extent of AI assistance accurately when necessary.

AI painting

SeaWhale AI offers text-to-image, image-to-image, image extraction, and a variety of other tools for image creation.

• The prompt should specify the subject, scene, composition, and style.
• Check the copyright of the reference image before generating an image from it.
• For product images, it is necessary to check the shape, text, and brand logos.
• The character image check verifies the hands, face, and portrait authorization.
• Confirm the licenses for the models and materials to be used before commercial use.
• When published publicly, indicate that it was generated by AI in accordance with the rules.

Skills and MCP

Skills encapsulates goals, data sources, and output formats into reusable workflows.

MCP is used to connect external tools with real-time data, so that AI is not reliant solely on the knowledge already present in the models.

• Select the Skill that corresponds to the task scenario.
• Confirm the data and external services that Skill will access.
• Only the minimum permissions required to complete the task are granted.
• Real-time data still requires verification of its source and timing.
• The number of skills available on the platform is constantly changing.
• High-risk operations should require manual confirmation.

Assistants and knowledge communities

The Assistant Center offers AI assistants designed based on roles and tasks, while the knowledge community contains product tutorials and public articles.

• View the assistant’s goals, limitations, and creator.
• Do not consider a description of assistants as a guarantee of competence.
• Community articles may contain personal opinions and promotional content.
• Professional advice requires verification by authoritative sources.
• Remove private and confidential information before publishing content publicly.

Client and platform

Web membership and computing power prices

The top-up page for regular users reveals information regarding VIP status and computing power, but the amount due is not displayed consistently on the page without logging in.

Do not estimate the VIP price based on empty pages or previous articles; the final price is determined by the account settlement page.

What is the SeaWhale AI API?

The developer API connects to models such as OpenAI, Claude, Gemini, and DeepSeek through a key and a unified interface.

• Compatible with the OpenAI SDK and chat completion interface.
• Provides a list of models, keys, and usage statistics.
• It supports billing for input, output, and caching across different models.
• Some models offer different service channels.
• The platform claims to have built-in fault detection and automatic rollback capabilities.
• Suitable for application, proxy, customer service, and content system integration.

API Integration Tutorial

1. Register a developer account and complete the security verification.
2. Top up a small test balance on the billing page.
3. Create an API key intended for use only with the current project.
4. Copy the exact model ID from the model list.
5. Point the client’s base address to Haijing AI’s unified interface.
6. First, use short prompts to test authentication, streaming output, and error handling.
7. Set timeout, retry, concurrency, and budget limits.
8. Log input, output, cache, and actual charges.
9. Complete content security and privacy assessments before going live.
10. Regularly rotate keys and delete credentials that are no longer in use.

API billing method

The API offers pre-payment and pay-as-you-go options; there is no fixed monthly subscription fee, and the balance is denominated in dollars.

• Input and output tokens typically have different unit prices.
• Cache reads and writes may be charged separately.
• Different service channels may have different prices.
• Longer contexts may enter higher price tiers.
• WeChat top-ups are settled at the exchange rate specified at the time of the order plus service fees.
• Invoices, bills, and real-time prices are the ultimate references.

Price of the model API representation

The following are the reference prices for dollars per million Tokens as displayed on the real-time price page as of August 31, 2026.

The price may include discounts and can vary depending on the model and channel; it is necessary to check the real-time model page and the bill before making a request.

API cost control

• Select the model with the lowest cost that meets the quality requirements.
• Limit the number of historical messages and irrelevant context.
• The stable system suggests using caching.
• Set a budget for each project and configure alert notifications.
• Record the additional costs incurred from retrying failed attempts.
• For batch tasks, start with a small-scale trial run before scaling up.

Top-up, refunds, and bills

The terms of service state that refunds are generally not available once the API top-up balance has been credited, and the free credits cannot be withdrawn or transferred.

• First, make a small deposit to verify the legitimacy of the business.
• Insufficient balance may lead to failed calls.
• The adjusted price applies to new calls.
• Objections to charges must be raised within 30 days after the fees are incurred.
• The actual amount paid in RMB is affected by the exchange rate and service fees.
• Companies should keep records of bills, invoices, and usage data.

API key security

• Do not write the keys in the frontend of web pages or mobile applications.
• Do not submit to a public GitHub repository.
• Use environment variables or a key management system.
• Create separate keys for different projects.
• Upon detecting an abnormal call, it is immediately revoked and replaced.
• The server imposes restrictions on sources, concurrency, and budget.

Privacy and third-party models

The aggregation platform forwards requests to the appropriate upstream models, and data processing is also governed by the rules of those upstream services.

• Do not submit passwords, keys, or unpadded personal data.
• The company first determines the data location and cross-border requirements.
• When a document is uploaded, only the content necessary to complete the task is retained.
• The retention and training policies for different models may vary.
• Establish models and supplier whitelists for sensitive industries.
• Do not record full prompts and keys in the logs.

GitHub and open source

At present, no official open-source repository that can be confirmed as belonging to the Haijing AI full platform has been found.

SeaWhale AI is compatible with the OpenAI SDK, but compatibility at the protocol level does not mean that the gateways, models, or the platform for ordinary users are open source.

Usage restrictions

• Web members, papers, and API usage have different billing systems.
• The aggregation model may become temporarily unavailable due to changes upstream.
• Model prices and discounts are adjusted according to market conditions.
• AI papers cannot replace genuine research and academic contributions.
• Third-party Skills and APIs may process user data.
• All important outputs require manual verification.

Frequently Asked Questions

Can Haijing AI be used for free?

You can experience some of the features and receive invitation rewards, but the specific number of free uses and the quota associated with the model are indicated on the account page.

What functions does Haijing AI support?

It supports multi-model dialogue, writing, academic document creation, paper assistance, AI painting, Skills, and MCP.

How much does a paper on Sea Whale AI cost?

The official website states that the first outline is available for free, while the cost of the main text starts at around 29 yuan; the final price is determined based on the number of words and selected options.

Does SeaWhale AI provide APIs?

Developers can use a unified interface that is compatible with the OpenAI protocol to invoke hundreds of models.

How is the Haijing AI API billed?

Prepaid pay-as-you-go is used; different prices may apply for input, output, caching, and service channels.

Can the balance in an API top-up be refunded?

It is generally not supported; the terms of service state that no refunds are given after the balance is credited, except in cases prescribed by law or approved by the platform.

Is SeaWhale AI an open-source platform?

No, there is currently no verifiable official repository containing the complete source code of the platform.