DEV Community

ModelFild
ModelFild

Posted on

ModelFold chat subscription: which plan, how quota works, and Claude Code or Codex

ModelFold now has a monthly subscription for chat models. This note only covers what the pricing page and the agent docs showed on 17 November 2026. Dollar prices, yuan prices, and discounts follow whatever the page shows when you check out.

What the subscription is

You pay by the month. While it is active, every chat model is included: Claude, GPT, Gemini, DeepSeek, Kimi, GLM, Grok, Qwen, and the rest of the chat catalog. Those calls do not spend credits.

It works in two places: the Playground on the site, and the API. The API path includes coding agents such as Claude Code and Codex.

Image, video, audio, and the other online tools still use credits. They are not part of this subscription.

Which of the three plans

Prices on the pricing page on 17 November 2026:

  • Pro: $20 per month (¥145). This is the base allowance. Up to 2 requests at once.
  • Max: $100 per month (¥720). 5× the allowance. Up to 4 requests at once.
  • Ultra: $200 per month (¥1440). 10× the allowance. Up to 8 requests at once.

All three plans include the same chat models. The difference is the allowance, and how many requests can run at the same time. If you use a coding agent every day, or you chat a lot, start by checking whether Pro is enough. If you need clearly more, take Max. If you need another step up, take Ultra.

At checkout you pick 1, 3, 6, or 12 months and pay that term once. The expiry date is on the subscription page, and renewal is done there too. Discounts shown on 17 November 2026: 5% off for 3 months, 10% off for 6 months (the 6-month price is 90% of the monthly price), and 15% off for 12 months. The page also shows 10% off the first month for a new subscriber. Those rates can change. The price calculated on the page at checkout is the one that applies.

While you are subscribed, credit top-ups are 10% off.

Invoices are available. Payment methods are the same set as credit top-ups, whichever ones the checkout page has open.

Where to buy:

How the allowance works

Two limits apply together: a rolling 5-hour window, and a weekly cap.

The page does not show a remaining request count. It shows the percent used, and when that limit resets. The 5-hour window starts with your next request. It does not start ticking empty from the moment you pay.

When a limit is used up, calls pause until the reset time shown on the page. The subscription page has a switch, "Use credits when allowance runs out". It is off by default. While it is off, a used-up allowance only waits. It does not start spending credits. Credits are used for the extra calls only if you turn that switch on.

If the page says one model family uses the allowance faster, follow that line. This article does not add a multiplier of its own.

Point Claude Code and Codex at it

The subscription uses your normal ModelFold API key. You do not create a second kind of key. Create one at https://modelfold.ai/account?tab=keys&utm_source=devto . Each key can turn "Use subscription" off on its own.

The Playground needs no extra setup. Sign in and chat.

The full steps for coding agents are in this guide: https://modelfold.ai/docs/ai-agent/overview?utm_source=devto

You can also install the skill from that guide and ask the agent to apply the steps. It needs Node.js 22.20 or newer:

npx skills add https://modelfold.ai --skill modelfold-chat-agents -g -a claude-code -a codex -a grok -y
Enter fullscreen mode Exit fullscreen mode

Put the key in the MODELFOLD_API_KEY environment variable, then tell the agent to configure Claude Code, or Codex, to use ModelFold.

If you configure it by hand, the addresses below are the ones in the docs. Take model IDs from the list endpoint. Do not fill them in from memory.

Claude Code:

export ANTHROPIC_BASE_URL=https://modelfold.ai/anthropic
export ANTHROPIC_AUTH_TOKEN=your_modelfold_key
Enter fullscreen mode Exit fullscreen mode

List models with that key at https://modelfold.ai/anthropic/v1/models . Stop the base URL at /anthropic. The client appends /v1/messages. Do not put a Bearer prefix inside the token. Set the main model and the three tier model variables the way the doc shows, using IDs from that list.

Codex belongs in your user config, not in a project file:

model = "a-slug-from-the-model-list"
model_provider = "ModelFold"

[model_providers.ModelFold]
name = "ModelFold"
base_url = "https://modelfold.ai/openai/v1"
env_key = "MODELFOLD_API_KEY"
wire_api = "responses"
Enter fullscreen mode Exit fullscreen mode

The model list is https://modelfold.ai/openai/v1/models . Do not put /responses on the base URL. Codex adds that path. env_key is the name of the environment variable. The key itself goes in MODELFOLD_API_KEY.

Subscription or pay as you go

Pay as you go means you add credits first and each call spends them. On the pricing page, failed jobs are not charged, and credits do not expire. That fits occasional use, and it fits image, video, and audio work.

The subscription fits people who use coding agents such as Claude Code or Codex often, or who chat a lot every day. With the default switch left off, chat spend for the term is the plan price, so you can see the cost before the month starts. If you do not use it much, do not subscribe.

You can keep both. Chat goes through the subscription. Image, video, and audio go through credits. Credit top-ups during a subscription are currently 10% off.

This article does not estimate a saving against any vendor's own price. The amount on the pricing page and the subscription page, at the moment you pay, is the price.

Top comments (0)