360 Smart Brain

What is 360 Smart Brain?

360 Smart Brain is a large-model technology and enterprise AI service system launched by the 360 Group. It first came to public attention in 2023 with its general-purpose chat functions and its own developed large models; currently, its focus has shifted to open platforms, multi-model APIs, enterprise knowledge bases, long-term memory, Agents, and private deployment options.

For ordinary users who need search functions, Q&A capabilities, and content creation tools, 360 currently promotes its NanoAI solution; developers and enterprises, on the other hand, continue to use the 360 Smart Brain open platform.

Nano AI makes use of 360’s proprietary intelligent brain models, as well as various models developed in China; however, it is not the same service as open platform accounts, API billing, or enterprise-specific customized solutions.

Current product positioning

• 360 Smart Brain Open Platform:It provides unified access to various text and multimodal models, offering an online interface, API keys, model routing, as well as usage billing.
• RAG knowledge base:It provides functions for parsing, segmentation, vector retrieval, and question answering for enterprise documents;
• Mem360 memory database:Add long-term memory across sessions for AI applications and Agents;
• Skill distribution:Switch prompts, models, and business capabilities based on intent;
• Enterprise privatization platform:Deploy models, knowledge bases, intelligent assistants, and security governance in the organizational environment;
• Nano AI:A portal for multi-model search, creation, and intelligent assistance tailored for individual users.

Multi-model open platform

The 360 Intelligence Brain Open Platform is based on the approach of \"single access, coverage of all models, and intelligent scheduling.\" The Model Plaza includes not only models developed or customized by 360 itself, but also models from various other providers such as Tongyi Qianwen, DeepSeek, Zhipu GLM, MiniMax, Kimi, and DouBao.

Developers can compare context, input/output costs, reasoning capabilities, supported parameters, speed, and recent availability.

The platform selects the appropriate supplier based on the scheduling strategy, and it automatically switches to an alternative provider when one of them becomes unavailable. The same logical model can have multiple actual suppliers, and the final bill is calculated based on the actual model that was used as well as the input and output tokens.

For applications that require strong consistency, it is necessary to lock down the supplier, model version, and parameters; one cannot assume that intelligent routing will produce identical results each time.

API compatibility and invocation

The platform offers a calling method based on API Keys; the base address uses 360’s unified API domain name, and text-based conversations are compatible with OpenAI Chat Completions. Some models also support OpenAI Responses, Anthropic Messages, or Google VertexAI formats – the specific capabilities are indicated on the model’s details page.

The API Key must be stored on the server, in a key management system, or within secure configuration settings; it cannot be embedded in web pages, clients, or public repositories. In a production environment, it is necessary to implement mechanisms for timeout handling, retry attempts, concurrency control, budget management, content filtering, and log anonymization. Additionally, records of the models and suppliers should be kept in order to track any changes in performance.

Model Plaza and prices

The platform displays the cost per million tokens for inputs and outputs separately for each model; there is no single uniform price that applies to all models. Examples from the Model Plaza as of July 2026 include:

The number of model squares, the available suppliers, as well as discounts and prices change in real time. Some models are designated for specific internal operations or testing purposes, and ordinary developers may not have access to them. The examples provided in the catalog are intended solely to help understand the billing mechanism; purchases and budgeting must be based on the information displayed in the console at that time.

360 Smart Brain RAG Knowledge Base

The RAG knowledge base analyzes internal corporate documents, emails, wikis, and other materials to create searchable indexes; before providing a response, it looks for relevant excerpts to supplement the private knowledge that the model is not aware of. It is suitable for use in internal Q&A systems, customer service, virtual assistants, document reading, smart hardware, and business assistants.

The current official pricing is 0.6 yuan per GB per day for storing data in the knowledge base, with each user entitled to 1 GB of free storage space. The cost for retrieving and uploading documents is based on the actual number of tokens used, at 0.5 yuan per million tokens.

For privatization, large-scale processing, or specialized functions, separate negotiations are required.

RAG can reduce unfounded responses, but it cannot eliminate hallucinations. Errors in document parsing, permission settings, outdated content, and omissions in the retrieval process can all affect the quality of the answers; therefore, companies should keep track of the source snippets, implement access controls, note the update dates, and conduct manual verification.

Mem360 long-term memory

Mem360 is a plug-in semantic memory layer developed by 360 Intelligence Brain for AI applications and agents. It enables the storage of user preferences, past tasks, customer service progress, and business context, and allows relevant memories to be retrieved based on the current issue, thereby reducing the overhead associated with including all historical data in the context.

The official documentation states that the memory database is available for free on a temporary basis, but the use of large models within the memory processing workflow is still billed based on the API Key provided by the user. This temporary free access is not a permanent offer; future pricing will be determined as indicated in the control panel.

Long-term memory involves privacy concerns and the accumulation of errors. Applications must allow users to view, modify, and delete memories, distinguish between facts, deductions, and temporary preferences, and limit the duration for which sensitive information is stored as well as the scope of access to it.

Skill distribution

Skill distribution allows for switching the system prompts or models based on the user’s intentions – for example, directing queries to customer service, virtual assistants, code, or industry-specific tools. Switching models is not necessary;

Replacing only the prompt words can also enable the development of different skills.

If another model is used after a hit, the charge will be based on the model that was switched to and the actual number of tokens.

Automatic routing may misinterpret intentions or route sensitive requests to tools with higher permissions. Before going live, it is necessary to evaluate the accuracy of classification, the fallback logic, permission boundaries, as well as peak RPM/TPM values, and to provide a manual entry point for critical operations.

Enterprise privatization and AI assistants

The 360 Smart Brain enterprise platform offers model integration, a knowledge base, file collaboration tools, an intelligent assistant, agents, and security management functions; it is capable of monitoring updates to corporate systems and synchronizing new knowledge into the knowledge base. Solutions intended for government agencies, enterprises, and large organizations usually require an appointment for a demonstration and a separate quote.

Privatization allows for control over data boundaries and network environments, but it does not mean zero risk or no need for maintenance. Customers still need to manage computing resources, model versions, patches, backups, permissions, logs, disaster recovery measures, and content security; they also need to determine whether third-party models are still accessed through external APIs.

The relationship between nano AI and 360 Smart Brain

NanoAI is a personal AI search and creation tool that 360 has been promoting intensively since 2024; it integrates 360 Smart Brain as well as various models from within China, and offers functions such as searching, answering questions, writing, as well as handling images and videos. The NanoAI assistant built into the 360 browser can also assist with reading and responding in the webpage sidebar, search results, and on social media platforms.

Therefore, ordinary users looking for \"360 Smart Brain Chat\" are better suited to use NanoAI; developers who need APIs, RAG, memory functions, or private solutions should turn to the 360 Smart Brain Open Platform.

The old chat entry may still exist, but the catalog should not describe the single chatting feature from 2023 as representing all the current products.

Open-source models and GitHub

On Hugging Face and ModelScope, 360 has released 360Zhinao, 360Zhinao2, search vector models, as well as vision research models under the official accounts qihoo360 or 360zhinao.

360Zhinao2 includes Chat versions with 7B Base, 4K, 32K, and 360K contexts, and offers BF16 as well as some Int4 weights, along with Transformers, vLLM, tuning options, and web demonstration instructions.

The source code of the 360Zhinao2 warehouse is labeled as Apache 2.0, and it allows free commercial use without the need for any special requests. Although the source code of the earlier 360Zhinao-search model is also labeled as Apache 2.0, there are separate terms or specific licensing conditions regarding its use in commercial purposes; therefore, it is necessary to check each warehouse individually.

The term \"open weight\" does not mean that the 360 Smart Brain open platform, NanoAI, RAG knowledge base, Mem360, or the enterprise control panel are open-source. The 360CVGroup repository on GitHub is primarily used to publish research projects in the field of computer vision; it does not contain the source code for a complete commercial platform.

360 Smart Brain User Guide

Complete a basic task.

1. Register 360 Smart Brain and create an API Key intended solely for the testing environment;
2. Select a model based on input type, context, quality, speed, and price;
3. First, invoke the current product positioning to complete the minimum request and check the returned structure;
4. Then use the multi-model open platform to test stream output, parameters, and abnormal response;
5. Record Tokens, number of calls, latency, error rate, and cost per call;
6. Move the key to the server-side key manager before integrating it into the actual application;

Create reusable professional workflows

1. Different keys and quotas are used for development, testing, and production environments;
2. A representative evaluation set is established based on the current product positioning, the multi-model open platform, and the 360 Smart Brain RAG knowledge base.
3. Set timeout, concurrency, retry, throttling, and budget limits;
4. Perform checks on the output regarding facts, security, format, and sensitive information;
5. Monitor changes in model version, price, latency, and failure rate;
6. Prepare plans for downgrading the model, implementing circuit breaking, and taking manual control;

Data security and usage restrictions

• The platform’s terms of service clearly state that the outputs of AI systems may be inaccurate, and therefore they should not be used as a substitute for professional judgment in high-risk areas such as medicine, law, finance, and security.
• RAG responses should indicate the sources of the information; online data must have their dates verified, and codes as well as Agent actions need to be tested and approved by humans.
• Before uploading corporate data, it is necessary to classify the data, remove sensitive information from it, and assign appropriate permissions.
• Avoid storing keys and full personal information in API logs;
• Nano AI and browser extensions may process web page content, so it is necessary to check the required permissions and privacy settings before installing them.

Which users is it suitable for?

• Developers who wish to compare and invoke multiple models through an API;
• An AI application team that requires intelligent routing, supplier fallback options, and unified billing;
• Establishing an organization for document Q&A and knowledge bases in construction companies;
• A product team that adds long-term memory capabilities for customer service, personal assistants, and agents;
• Government and enterprise clients that require privatization as well as governance for permissions and security;
• Researchers and developers who wish to deploy the 360Zhinao2 open model locally.

Frequently Asked Questions

Can 360 Smart Brain still be used?

Yes. Currently, services are provided primarily through open platforms, RAG knowledge bases, Mem360, and enterprise-specific private solutions.

The entry point for individual search and creation is primarily nano AI.

Is 360 Smart Brain free?

The open platform model typically charges based on the number of tokens; certain features, free quotas, and exemptions offered by Mem360 may vary. RAG provides 1GB of free storage per user, with additional storage costing a fee based on usage.

Are APIs provided?

Available. It is possible to create API keys and use interfaces that are compatible with protocols such as OpenAI; Model Plaza charges different fees based on the respective suppliers.

How much is the RAG knowledge base?

The current storage cost is 0.6 yuan per GB per day, with 1 GB of free storage; the cost for retrieval and document uploading is 0.5 yuan per million tokens. Separate quotes are provided for private deployment and large-scale use cases.

Is 360 Smart Brain the same as Nano AI?

It’s not exactly the same. NanoAI is a product designed for individual use in searching and creating content, while 360 Smart Brain refers more to underlying models, developer APIs, and enterprise-level solutions.

Is 360 Smart Brain open-source?

The entire platform is not open-source, but the weights and code for certain models such as 360Zhinao2 are made available. Whether commercial use is permitted depends on the license associated with each repository and model.