Qwen3-Next-80B-A3B Instruct is a highly efficient instruction-tuned large language model designed for fast, stable responses with ultra-long context handling and high throughput. It activates only a small portion of its 80 billion parameters to achieve significant improvements in speed and cost-efficiency without sacrificing performance in reasoning, code generation, and other complex tasks.
Qwen3-Next-80B-A3B Instruct activates only about 3 billion parameters out of 80 billion during inference, making it much faster and cheaper to run — about 10 times faster and more cost-efficient compared to the earlier Qwen3-32B model. It delivers over 10 times higher throughput on long contexts of 32K tokens or more. The model supports flexible deployment options including serverless, on-demand dedicated, and monthly reserved hosting. It is compatible with SGLang and vLLM for deployment with advanced mult-token prediction capabilities, ensuring efficient and scalable usage.
Input: $0.195
Output: $1.56
vs Qwen3-235B: The 80B A3B model matches or closely approaches the flagship 235B in reasoning and code tasks but is much more efficient, activating fewer parameters for faster, cheaper inference.
vs GPT-4.1: Qwen3-Next offers comparable instruction-following and long-context capabilities, with an edge in throughput and token window size, making it suitable for extensive document comprehension.
vs Claude 4.1 Opus: Qwen3-Next provides superior performance in multi-turn dialogues and agentic workflows, with more deterministic outputs on very long contexts compared to Claude’s conversational strengths.
vs Gemini 2.5 Flash: Qwen3-Next shows better scaling in ultra-long context handling and multi-token prediction efficiency, giving it an advantage in processing complex, multi-step reasoning tasks.
Measured on the AI/ML API gateway across live traffic. Figures are medians unless stated. Last updated 2026-10-12.
import requests r = requests.post( "https://api.aimlapi.com/v1/chat/completions", headers={"Authorization": "Bearer " + AIMLAPI_KEY}, json={ "model": "alibaba/qwen3-next-80b-a3b-instruct", "messages": [ { "role": "user", "content": "Hello!" } ] }, ) print(r.json())
const r = await fetch("https://api.aimlapi.com/v1/chat/completions", { method: "POST", headers: { Authorization: `Bearer ${process.env.AIMLAPI_KEY}`, "Content-Type": "application/json", }, body: JSON.stringify({ "model": "alibaba/qwen3-next-80b-a3b-instruct", "messages": [ { "role": "user", "content": "Hello!" } ] }), }); console.log(await r.json());
curl -X POST https://api.aimlapi.com/v1/chat/completions \ -H "Authorization: Bearer $AIMLAPI_KEY" \ -H "Content-Type: application/json" \ -d '{"model":"alibaba/qwen3-next-80b-a3b-instruct","messages":[{"role":"user","content":"Hello!"}]}'
OpenAI-compatible — swap the base URL and it works with your existing SDK.
| Type | Price |
|---|---|
| Input | |
| Output | |
| Benchmark | Score | What it measures | Source | Retrieved |
|---|---|---|---|---|
| MMLU-Pro | 80.6% | Multi-discipline knowledge + reasoning (harder MMLU) | Source | July 12, 2026 |
| LiveCodeBench | 56.6% | Contamination-free competitive programming problems | Source | July 12, 2026 |
| AIME 2025 | 69.5% | Competition mathematics (AIME), 2025 | Source | July 12, 2026 |
| GPQA Diamond | 72.9% | Google-proof graduate science questions (hardest subset) | Source | July 12, 2026 |
| Intelligence | 9.6 | Composite score across standardised reasoning, knowledge and problem-solving evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Math | 66.3 | Composite score across standardised mathematics evaluations, measured independently by Artificial Analysis | Source | September 12, 2026 |
| Model | Input | Output | Context | Best for |
|---|---|---|---|---|
Qwen3-Next-80B-A3B Instruct This page | Coding + agents | |||
| Reasoning + agents | ||||
| Balanced coding + agents | ||||
| Long-context, multimodal & agentic workflows | ||||
| Reasoning + agents |
Qwen3-Next-80B-A3B Instruct has a 129,024 tokens context window and can return up to 16,384 tokens.
Qwen3-Next-80B-A3B Instruct takes text as input and returns text.
Use alibaba/qwen3-next-80b-a3b-instruct as the model id. Requests go to https://api.aimlapi.com/v1/chat/completions.
Qwen3-Next-80B-A3B Instruct became available on September 30, 2025.
Qwen3-Next-80B-A3B Instruct is priced at input $0.195 / 1M tokens, output $1.56 / 1M tokens.
Yes, Qwen3-Next-80B-A3B Instruct can stream responses as they are generated.
Qwen3-Next-80B-A3B Instruct was built by Alibaba Cloud.
It supports parallel tool calls, structured output, function calling, and web search.
