NextHub aggregates access to mainstream models such as Claude, GPT and Gemini behind a single gateway: metered by token, traceable per request, and Chat Completions, Messages and Responses all served through the same endpoint. Which models and groups you can use depends on your account permissions and console configuration.
Four typical setups, each with one main problem and one next step; per-tool configuration is covered page by page in the docs.
For users of Claude Code, Codex, OpenCode, Kilocode, Zed and similar tools.
Every tool has its own address, protocol, model names and configuration, so switching tools means setting everything up again.
One account, one API Key, one entry point; follow the tutorial for the tool you use. This page does not repeat the tutorial contents.
For developers already using the OpenAI or Anthropic SDK.
Addresses, authentication and billing are split across providers, so accounts and costs are hard to consolidate.
All three published protocols are supported: adjust the Base URL, key and model configuration to connect; mainstream SDKs keep working the way they already do. The quickstart below gives a minimal example for each protocol.
For developers choosing between GPT, Claude, Gemini and other authorized models.
Model IDs, protocols, channels and cost bases all differ, so integrating each one is expensive to maintain.
One entry point, groups that define what you can call, and per-request usage records; the group determines the routing channel and billing multiplier, while load balancing is handled on the platform side.
For teams where several people, projects or environments share the API.
One master key shared by everyone: no spending boundary, and usage cannot be traced back to a person or project.
Set a spending limit, expiry, IP allow / deny rules and 5-hour / daily / weekly rate limits per API Key.
From account pool and group routing to API Key governance: all six items below map to capabilities listed in the published terms and the docs, and quote no figures without a stated basis.
Aggregates access to model families such as Claude, GPT and Gemini behind one endpoint: Chat Completions, Messages and Responses are all available, and replacing the base URL plus key is enough to connect.
→Upstream is provided through official account pools; requests pick a route by account pool and group, while load balancing and request scheduling are handled on the platform side.
→Metered by token and settled by group multiplier; every request's token usage and charge is recorded and can be checked line by line in the console.
→The group you choose when creating an API Key determines which models you can call and which routing channel is used; the specific upstream choice follows the platform's routing rules.
→Set a spending limit, expiry, IP allow / deny rules and 5-hour / daily / weekly rate limits per API Key.
→Official account pools connect directly to upstream official APIs; the platform runs health checks, error classification and risk-control isolation, and keeps billing, audit and risk-control records.
→The previous section covered what exists; this one covers how those capabilities work together within a single call.
Your SDK, CLI or IDE extension sends the request to the NextHub endpoint.
It lands on the Chat Completions, Messages or Responses endpoint, with the protocol format adapted on the gateway side.
The API Key is verified and that key's spending limit, expiry, IP rules and rate-limit windows are applied.
The group chosen for the API Key and the platform's routing rules decide which upstream route and model this request uses.
The result goes back to the client while token usage and billing details are recorded for this request.
Where step 4 leads is decided by platform policy. Under the published Service-Specific Terms, requests may be routed to different upstream resources based on model, account pool, region, balance, concurrency, rate limits, health status, sticky sessions, failover or NextHub policies — so whether a route is switched at any step depends on conditions at the time and it is not guaranteed to happen. See Service-Specific Terms.
The group you pick when creating an API Key decides three things: which models that key can call, which upstream route the request takes, and which multiplier is billed.
Exclusive groups for image generation and other services will only be displayed after logging into the console and enabling corresponding permissions; therefore they are not listed above; actual available models, permissions, and billing rates are subject to the console and Model Plaza.
Model Plaza: full model list and the prices you actually pay → Console →
All three protocols are supported — Chat Completions, Messages and Responses. Swap the endpoint, add your key, and NextHub handles the rest.
Register a NextHub account and open your personal console.
In the console, choose a model group, set the limit, expiry and rate limits as needed, and copy the platform key.
Pick any of the three protocols and replace the base URL and key; mainstream clients and the official SDKs keep working the same way.
# OpenAI-compatible · swap base URL and key to call curl https://api.nexthub.cc/v1/chat/completions \ -H "Authorization: Bearer sk-nx-your-key" \ -H "Content-Type: application/json" \ -d '{ "model": "gpt-6-astra", "messages": [{"role": "user", "content": "Hello, NextHub!"}] }'
# Anthropic-compatible · just swap base URL and key curl https://api.nexthub.cc/v1/messages \ -H "x-api-key: sk-nx-your-key" \ -H "anthropic-version: 2023-06-01" \ -H "Content-Type: application/json" \ -d '{ "model": "claude-fable-5-1", "max_tokens": 1024, "messages": [{"role": "user", "content": "Hello, NextHub!"}] }'
# OpenAI Responses protocol curl https://api.nexthub.cc/v1/responses \ -H "Authorization: Bearer sk-nx-your-key" \ -H "Content-Type: application/json" \ -d '{ "model": "gpt-6-astra", "input": "Hello, NextHub!" }'
Each entry points to that tool's own tutorial page in the docs; follow that tutorial for the configuration steps.
These tools are products of their respective vendors and NextHub has no official partnership with them — the entries simply link to tutorials that describe how to connect them to NextHub. CC-Switch is a third-party configuration helper, not a NextHub product.
Billing follows actual usage; the three rows below cover how individuals and teams pay, the top-up basis, and enterprise customization.
Settled by actual token usage × the multiplier of the group you select, with usage and charge recorded for every request, checkable line by line in the console.
Minimum top-up of 5 USD per transaction, no fees, Stripe (Visa / MasterCard / Alipay / AMEX) and crypto (USDT / USDC) supported, topped up at the live exchange rate; the console's actual payment page prevails.
Dedicated channel, custom quota and human support; the exact scope and settlement follow the order or a written agreement between the parties.
This page explains the billing basis only and does not reproduce the full price list. For the latest prices per group and the complete model list, see the Model Plaza.
NextHub publicly provides current mainstream versions of the Claude, GPT, Gemini and Grok model families and of Chinese open-source models such as DeepSeek, GLM and Step. Available models follow upstream updates. Image generation and other services requiring exclusive groups are only visible after logging into the console and enabling corresponding permissions. In other words, which models you can call is determined jointly by account permissions and selected groups; the complete list is subject to Model Plaza and console.
All three mainstream protocols are supported: OpenAI Chat Completions (/v1/chat/completions), Anthropic Messages (/v1/messages) and OpenAI Responses (/v1/responses), with https://api.nexthub.cc as the shared endpoint. Which protocol a given model accepts depends on the models offered by that group and on the docs.
Yes. Point the base URL at NextHub and supply your platform key, and mainstream SDKs and clients keep working the same way — you change only the base URL and key, without rewriting application code. For the exact base URL see "Does the Base URL need a version path?" below.
Per-model unit price × tokens actually consumed, where paid = official price × the multiplier of the selected group. Token usage and charges are recorded for every request and can be checked line by line in the console; group multipliers are described in the Model Plaza.
To handle authentication, forwarding, billing, risk control, troubleshooting and compliance, the platform processes the necessary request metadata: account information, API Key identifiers, IP and device information, routing information, model names, token usage, error information, order information and logs. Once a request is forwarded upstream, the related inputs, outputs and logs may be processed by that provider under its own terms; please do not submit unnecessary sensitive personal information, keys or regulated data. The complete position is set out in the Privacy Policy and in the data and log boundaries section of the Service-Specific Terms.
Start with the troubleshooting section of the docs; if that does not help, the FAQ and contact page lists the ticket entry point and administrator contact details.
It depends on the models you intend to use: select the claude code group for Claude (no client restriction); select the Claude Max group for Claude if you want a more stable pool; select the Claude-AWSB group for Claude via Amazon Bedrock; select the GPT-Azure group for GPT / Codex series; choose the GPT-Pro group for lower pricing; for promo pricing, choose the GPT-offers group (not recommended as a main coding option); select between Gemini-aistudio (Economy) and Gemini-Vertex (Official Stable) for Gemini based on budget; select the Grok group for Grok series; select the 开源模型组 group for DeepSeek, GLM, Step and other open-source models. The “Choosing models and groups” section above on this page lists each group with its purpose, current multiplier and restrictions; group names and multipliers are sourced from this page's snapshot; you can also directly view Groups and Multipliers Guide.
It depends on the protocol and the tool — there is no address that works for every case. OpenAI-style interfaces (Chat Completions, Responses) use https://api.nexthub.cc/v1, with /v1; Anthropic-style usage (the Messages protocol, Claude Code and the Anthropic SDK) takes https://api.nexthub.cc as the base URL, without /v1 — the SDK appends the path automatically, and the full endpoint for direct requests is https://api.nexthub.cc/v1/messages. OpenAI-compatible clients such as Zed and Kilocode take https://api.nexthub.cc/v1. Follow the tutorial for the tool you use.
A 429 can come from different causes, so check each in turn: first whether your account quota and balance are sufficient, since insufficient quota or balance is rejected outright; if quota and balance are fine and it still fires often, you have most likely hit the rate limit of the API Key or of its group (5-hour / daily / weekly windows), so retry later or adjust that key's rate settings. Also confirm that the model you are calling belongs to the group selected for that key.
The Model Plaza is maintained directly by the platform backend; groups and prices are synchronized automatically as backend configurations and upstream official prices change, so what you see when you open the page is always the latest data. What you are actually charged still follows the console record at request settlement time; if you have questions about a bill, contact an administrator with the request ID.
Register to create your first API Key and choose the model group you want to use.