---
title: "Qwen API — 變體、價格與 model skill | RunAPI"
url: "https://runapi.ai/zh-TW/models/qwen.md"
canonical: "https://runapi.ai/zh-TW/models/qwen.md"
locale: "zh-TW"
model: "Qwen"
provider: "Alibaba"
modality: "text"
variant_count: 1
price_from_cents: 1
---

# Qwen API

Alibaba Qwen API access via RunAPI — Apache-2.0 ultra-sparse MoE with 262K context, 80B total / 3.9B active.

**Provider:** Alibaba
**Modality:** Text
**Catalog:** 1 endpoints

Qwen is Alibaba's Apache-2.0 family of language models. qwen3-next-80b-a3b-instruct uses a hybrid attention architecture (DeltaNet linear + GQA) with an ultra-sparse MoE — 80B total parameters, only ~3.9B active per token across 512 experts. It delivers 262K native context (extendable to 1M) and matches Qwen3-235B-A22B on coding and conversational benchmarks while using 7× fewer active parameters and 10× higher throughput. Available through RunAPI with one key and per-token billing.

## Variants

Single-SKU line. The SDK model ID is `qwen3-next-80b-a3b-instruct` and all usage details are documented here.

## Pricing

| Endpoint | Pricing | Billing |
|---|---|---|
| `chat_completion` | $0.010 | 1K tokens |

## Spec sheet

| Field | Value |
|---|---|
| Model ID | `qwen3-next-80b-a3b-instruct` |
| Provider | Alibaba |
| Modality | text |
| Task type | synchronous |
| Billing unit | 1K tokens |
| API endpoint | `/v1/chat/completions` |


## API endpoints

Base URL: `https://runapi.ai`

- `POST /v1/chat/completions`

Use the OpenAI or Anthropic SDK with your RunAPI API key. No extra SDK required.

## Context

Qwen models from Alibaba are Apache-2.0 ultra-sparse MoE LLMs with 262K native context. qwen3-next-80b-a3b-instruct matches models with 7× more active parameters on LiveCodeBench while running at 10× throughput. Through RunAPI they share a single API key with pay-as-you-go token billing, callable from the OpenAI Chat Completions, OpenAI Responses, and Anthropic Messages surfaces. These are Qwen text models, distinct from the Qwen 2 image line.

## FAQ

### 我應該先從哪個版本開始？

先選擇符合你品質標準中最便宜的版本。大多數團隊會先用快速版本，之後再升級到專業版用於正式上線。

### 有免費方案嗎？

新帳戶可在每個模型上免費使用首次呼叫。之後則按次計費。

### 你們支援串流結果嗎？

只要該功能可用，RunAPI 會提供端到端串流。

### 失敗的請求如何計費？

生成失敗不會收費。

### 輸出結果有快取嗎？

生成結果會儲存，並可透過任務 ID 取回。輸入內容不會快取。

### 可以商業使用嗎？

可以——除非模型授權明確限制，否則每個版本都包含商業使用權；若有例外，會在版本頁面標示。

### 速率限制怎麼算？

每個金鑰的速率限制會依使用等級而提升。最新限制請參考定價頁面。

### 如果遇到問題，要去哪裡回報？

請在公開的 GitHub repo 開 issue，或寄信給支援。