# Qwen 3.6 Plus

> Alibaba's newest flagship. #1 on Kyma.

Human version: https://kymaapi.com/models/qwen-3.6-plus
Live JSON: `GET https://kymaapi.com/v1/models` (no auth required)

## Facts

- **Model ID**: `qwen-3.6-plus` — pass this as `model` in the request body
- **Creator**: Alibaba
- **Released**: 2026-04-02
- **Context window**: 131K tokens
- **Max output**: 33K tokens per response — a hard ceiling, not a default
- **Price**: $0.4388 in / $2.633 out per 1M, cached input 10%
- **Capabilities**: tools, vision, reasoning, caching
- **Alias**: `best` resolves to this model

## Call it

```bash
curl https://kymaapi.com/v1/chat/completions \
  -H "Authorization: Bearer $KYMA_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "qwen-3.6-plus", "messages": [{"role": "user", "content": "Hello"}]}'
```

## Positioning

The most-used model on Kyma and the one the `best` alias points to — frontier-grade reasoning at open-model pricing.

## About

Qwen 3.6 Plus is the flagship tier of Alibaba's Qwen family, built for general reasoning, coding, and multilingual work. It sits in Kyma's frontier-open quality tier: the class of open models that holds its own against closed flagships on everyday production tasks.

On Kyma it serves more production traffic than any other model. Requests sent with the `best` alias resolve here, and every call gets the platform's automatic failover — if a serving path degrades, your request is rerouted before you notice. Prompt caching is fully supported, so agents that resend long system prompts pay 90% less on the repeated part.

The 131K-token context window (roughly 190 pages of text) with function calling, structured outputs, and extended reasoning makes it a complete engine for agent workloads — not just chat.

## Use cases

- **Agent backbones** — Tool calling plus structured outputs make it a dependable engine for multi-step agents and function-calling pipelines.
- **Coding assistants** — One of the most common workloads it serves on Kyma — code generation, review, and refactoring inside coding agents.
- **Long-document analysis** — The 131K context fits whole codebases, contracts, or research papers in a single request.
- **Multilingual work** — First-class quality beyond English — the Qwen family is known for strong performance across Asian languages.
- **Reasoning-heavy tasks** — Extended reasoning mode handles math, analysis, and planning problems that trip up lighter models.

## Not ideal for

Image inputs (it's text-only — pick a vision model) or latency-critical UX, where a fast-tier model feels snappier for less money.

## Pick something else when

- You need the strongest tool-heavy agent behavior → `kimi-k2.6`
- You need deeper reasoning than speed → `deepseek-r1`
- You need lower latency for coding loops → `qwen-3-32b`
- You need 1M context → `gemini-2.5-flash`

## See also

- All models: https://kymaapi.com/models.md
- Pricing: https://kymaapi.com/pricing.md
- Other models by Alibaba: https://kymaapi.com/models?q=Alibaba