> ## Documentation Index
> Fetch the complete documentation index at: https://docs.morphllm.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Ultra plan

> Unlimited Kimi K3 and DeepSeek V4.1 Flash for $200 a month, subject to rate limits

The Ultra plan is \$200 a month. It removes the token meter on Kimi K3 and DeepSeek V4.1 Flash.

> [!WARNING]
>
> Unlimited does not mean unthrottled. Each subscription has per-minute limits on uncached input and output tokens for each model, and requests can return `429` when Morph is at capacity. See [Rate limits](#rate-limits).

## What the plan covers

| Model | API ID | On the plan |
| - | - | - |
| Kimi K3 | `morph-kimik3` | No token meter |
| Kimi K3.1 | `morph-kimik31` | No token meter |
| DeepSeek V4.1 Flash | `morph-dsv41flash` | No token meter |
| DeepSeek V4.1 Flash 1017 | `morph-dsv41flash-1017` | No token meter |
| Every other model, including `morph-kimik3-fast` and `morph-dsv41flash-abliterated` | | Per token at list price |

The plan includes no credits. Other models draw down any credits you have bought, then bill to your card at [list price](https://www.morphllm.com/pricing). There is no monthly token cap on the included models.

## Rate limits

Two kinds of limit apply to the included models.

**Per-minute rate limits.** Each subscription has one rate limit on uncached input tokens and another on output tokens for each model, so no single account can crowd out the rest. There is no limit on total tokens, only on how fast you use them.

| Model | Uncached input tokens per minute | Output tokens per minute |
| - | - | - |
| Kimi K3 and K3.1 | 24,290 | 6,162 |
| DeepSeek V4.1 Flash, both versions | 701,079 | 76,559 |

Cached input does not count toward either limit. A request bigger than a limit still runs, and your next request waits until that limit has caught up. Output counts when a response finishes. The `429` states the limit you reached, and `Retry-After` says how long to wait. Kimi K3 and K3.1 share one set of limits, as do the two DeepSeek V4.1 Flash versions.

**Capacity.** The plan does not reserve capacity. When Morph is at capacity, requests can return `429` until capacity frees up.

Handle both the same way: retry with exponential backoff, and honor `Retry-After` when it is present. The OpenAI and Anthropic SDKs do this for you.

```typescript theme={null}
import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.MORPH_API_KEY,
  baseURL: "https://api.morphllm.com/v1",
  maxRetries: 5,
});
```

For reserved capacity with an availability SLA, use [dedicated endpoints](https://docs.morphllm.com/dedicated-endpoints.md).

## Get the plan

1. Open [Billing](https://www.morphllm.com/dashboard/billing?plan=ultra).
2. Choose **Ultra** under **Subscription**.
3. Complete checkout.

The plan is for individual accounts and covers your personal API keys. Organizations cannot subscribe, and organization API keys are not covered.

If you already have a subscription, the new plan replaces it. Remaining subscription credits are not carried over.

## Use it with a coding agent

After checkout, [Set up Morph Ultra](https://www.morphllm.com/dashboard/ultra/setup) has a copy-paste prompt for Claude Code, Codex, Pi, OpenCode, T3 Code, and OMP. Paste the prompt into any coding agent and it configures that tool for you.

To configure Claude Code by hand, point it at Morph with the included models:

```bash theme={null}
export ANTHROPIC_BASE_URL="https://api.morphllm.com"
export ANTHROPIC_AUTH_TOKEN="YOUR_MORPH_API_KEY"
export ANTHROPIC_MODEL="morph-kimik3"
export ANTHROPIC_SMALL_FAST_MODEL="morph-dsv41flash"
claude
```

See [Claude Code & Codex](https://docs.morphllm.com/guides/coding-agents.md) for persistent settings and Codex setup.

## Cancel

Cancel from [Billing](https://www.morphllm.com/dashboard/billing). After the plan ends, usage on these models bills per token.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.