Kimi Coding Plan: Which Option Should You Choose?

Three ways to buy Kimi access

Start with where you want to use Kimi. A coding-agent connection, an official membership and a metered API account have different credentials and allowances.

On smaller screens, scroll the tables sideways to compare all columns.

Our model catalog lists included gateway models and rates. Current purchase options are shown in the Multi-Model group on Pricing.

For a developer who uses Kimi for some tasks and DeepSeek or GLM for others, the practical difference is how those services are managed. Our plan keeps the listed models behind one gateway key. An official Kimi membership serves a different need: access tied to your Kimi account and its benefits.

What does our plan include for Kimi K3.1 Pro and High Speed?

Our Multi-Model plan lists Kimi K3.1 Pro and Kimi K3.1 Pro High Speed. Use kimi-k3.1-pro for Pro at 15× or kimi-k3.1-pro-highspeed for High Speed at 20×. The existing kimi-k3 ID now selects Pro at 15× with your current key and connection settings.

Reasoning remains enabled on this route. Low reduces effort rather than turning reasoning off. Start with a representative task, inspect the result, then adjust effort if the additional reasoning is useful for your work.

For screenshot-based UI work, use the image input supported by your configured coding tool. Image understanding does not add video support, Kimi’s built-in browser or access to its other apps. Our gateway guides describe the connection and model settings for each documented tool.

How many Kimi K3.1 Pro tokens do the points cover?

Kimi K3.1 Pro uses 15 points per token; High Speed uses 20. The entry plan’s 60M-point 5-hour allowance therefore corresponds to 4M combined input and output tokens if used entirely on Pro, or 3M tokens on High Speed. It is not a 60M-token Kimi allowance.

Cached input counts once within input usage, and reasoning counts once within output usage. Your coding agent may resend project context and conversation history across several calls. Include those calls when estimating a session’s consumption.

M = 1,000,000; B = 1,000,000,000. Token equivalents assume the entire unused allowance is spent on Kimi K3.1 Pro and include both input and output. They are not separate Kimi balances. See Pricing for current stock, one-time purchase options and billing terms.

Every call counts toward both windows. You must have sufficient points in each; the two figures cannot be added together. The 7-day limit still applies after a 5-hour window resets. Switching models does not replenish either balance.

Example: Kimi and DeepSeek in the same session

Assume you begin with both entry-plan allowances unused. Within one 5-hour window, your agent sends the following hypothetical usage:

The remaining 20M points could cover another 1,333,333 Kimi K3.1 Pro tokens at 15 points each, provided the 7-day balance is also sufficient. This example illustrates accounting; it does not predict how many coding tasks you can complete.

How does official Kimi membership differ?

Official Kimi Code is a benefit within Kimi membership. For international customers, current Plus, Pro, Max and Ultra plans include coding access. New plans have a rolling 5-hour limit and a monthly total; legacy plans can also retain a weekly limit.

Check the account’s actual plan, rather than assuming all Kimi subscriptions use the same quota rules. Usage across logged-in devices and API keys draws on that account’s allowance. Eligible subscribers can also enable separately charged Extra Usage.

Those are official membership rules. Our Multi-Model plan has its own 5-hour and 7-day points balances. Kimi’s Extra Usage wallet and membership credits do not top up our gateway allowance.

If your main requirement is Kimi’s official Desktop, CLI, VS Code experience or other Kimi account benefits, assess the official membership on those terms. If you primarily want Kimi inside a coding workflow that also uses other model providers, compare our plan using the token equivalents above.

Choose a plan around your actual workload

• You switch between Kimi, DeepSeek and GLM. Compare Multi-Model tiers against your busiest 5-hour session and expected 7-day use. Included models share the balance, so calculate the whole workflow.
• Kimi is your main model for long coding sessions. Measure how quickly it consumes points before choosing a tier. A larger allowance may suit you, but compare official Kimi membership as well. One gateway key alone is not evidence of a lower cost.
• You are building a customer-facing application. Evaluate the official Kimi API Platform’s usage billing and service requirements. It is separate from Kimi membership and Kimi Code. Its token accounting distinguishes input, output and cache writes, so compare the full bill rather than output tokens alone.

Use a small set of real tasks for your comparison: a bug fix, a change across several files, or a UI change using a screenshot. Track accepted results and points spent, including failed attempts. The useful figure is the cost of a result you can use, not just the first response.

Also comparing GLM access? Our GLM Coding Plan guide explains its model-specific points rates and buying options.

Connect Kimi K3.1 Pro and High Speed in Claude Code or OpenCode

For our Multi-Model plan, use the delivered gateway key with the setup guide for your tool. A Kimi membership login, a Kimi API Platform key and our gateway key belong to different services.

1. Open the Multi-Model group on Pricing.Review the product’s stock, points limits and payment mode. The separate DeepSeek Coding Plan has its own allowance and does not include our Kimi route.
2. Use the delivered connection details.Our OpenAI-compatible API base is https://multi-api.deepseekv4pro.com/v1. Claude Code needs the corresponding Anthropic-compatible configuration from its guide.
3. Apply the tool-specific setup.Follow our Claude Code guide or OpenCode guide. Select kimi-k3.1-pro for Pro or kimi-k3.1-pro-highspeed for High Speed. Existing kimi-k3 settings continue to select Pro.
4. Verify the selected model and usage.Run a small task, confirm that the selected route is Kimi K3.1 Pro and review both points balances in the Dashboard before starting a longer session.

In Claude Code, choosing a GLM, DeepSeek or Kimi route changes the model answering through the gateway. It does not add Claude model access. If your workflow uses subagents or background models, review their configured routes too; changing the main model does not automatically change every separate alias.

Questions before purchasing