# Model APIs drift. Your app should not.

Markdown version of https://kymaapi.com, for agents and crawlers. Same content as the HTML page.

- Generated: 2026-10-10T00:08:32.298Z
- HTML page: https://kymaapi.com
- Model catalog: https://kymaapi.com/models.md
- Pricing: https://kymaapi.com/pricing.md

Pay for what you use · No card · One balance

Moving between models should not force endpoint rewrites, new billing setups, or separate reliability checks. Kyma keeps one API contract for chat, reasoning, image, video and audio so your app can ship and switch faster.

$0.50 signup credit, no card required. [Credit details](https://kymaapi.com/pricing#free-tier)

Know the cost before you build

## Real models, published prices.

106 models share one key and balance.

Swipe the table to compare all rates.

Published US dollar rates per one million tokens at the base context tier

| Model | Input / 1M | Cached input / 1M | Output / 1M |
| --- | --- | --- | --- |
| [Qwen 3.6 Plus](https://kymaapi.com/models/qwen-3.6-plus) | $0.454 | Not published | $2.724 |
| [DeepSeek V4 Flash](https://kymaapi.com/models/deepseek-v4-flash) | $0.1264 | $0.0243 | $0.2526 |
| [Gemini 3.7 Flash](https://kymaapi.com/models/gemini-3.7-flash) | $1.013 | $0.050625 | $5.063 |

81% less for cached input on DeepSeek V4 Flash.

[All prices →](https://kymaapi.com/pricing)

Know the reliability

## Measured uptime, model by model.

Automatic fallback keeps requests moving when a compatible route is available.

- [Qwen 3.6 Plus](https://kymaapi.com/models/qwen-3.6-plus) · 99.9% · 1,138 observations in the last 30 days
- [DeepSeek V4 Flash](https://kymaapi.com/models/deepseek-v4-flash) · 100% · 36,997 observations in the last 30 days
- [Gemini 3.7 Flash](https://kymaapi.com/models/gemini-3.7-flash) · 98.9% · 183 observations in the last 30 days

[How uptime is measured →](https://kymaapi.com/status#measurement)

## Your first request in three lines.

Use your existing OpenAI or Anthropic SDK.

**OpenAI SDK**

```
from openai import OpenAI
client = OpenAI(base_url="https://kymaapi.com/v1", api_key="YOUR_KYMA_KEY")
print(client.chat.completions.create(model="qwen-3.6-plus", messages=[{"role": "user", "content": "Hello!"}]).choices[0].message.content)
```

**Anthropic SDK**

```
from anthropic import Anthropic
client = Anthropic(base_url="https://kymaapi.com", api_key="YOUR_KYMA_KEY")
print(client.messages.create(model="qwen-3.6-plus", max_tokens=1024, messages=[{"role": "user", "content": "Hello!"}]).content[0].text)
```

OpenAI: `/v1/chat/completions`. Anthropic: `/v1/messages`. [SDK setup →](https://docs.kymaapi.com/quickstart)

Connect through compatible clients

- OpenClaw
- Cline
- Roo Code
- Cursor
- Claude Code
- OpenCode
- Aider
- Continue

One API contract across surfaces

## Ship the same app flow across chat, compare, and MCP.

Every screenshot below is a live Kyma surface. One endpoint and one balance carry the same product from model catalog to comparison to assistant setup.

Model catalog: one schema and one price table.

Compare: measured performance, not ad copy.

MCP setup: connect assistants without custom glue code.

## Choose with evidence

Compare models using customer usage and measured performance.

[Explore rankings →](https://kymaapi.com/rankings)

## Connect your AI assistant

Connect your assistant to the live catalog and chat tools through MCP.

[Set up MCP →](https://kymaapi.com/mcp)

https://mcp.kymaapi.com/mcp

## Start with one account.

Choose a model and build with Kyma API.

[Create account](https://kymaapi.com/signup)