qwen3-coder-next

1.3M Downloads Updated 3 months ago

Qwen3-Coder-Next is a coding-focused language model from Alibaba's Qwen team, optimized for agentic coding workflows and local development.

tools cloud

ollama run qwen3-coder-next

curl http://localhost:11434/api/chat \
  -d '{
    "model": "qwen3-coder-next",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

from ollama import chat

response = chat(
    model='qwen3-coder-next',
    messages=[{'role': 'user', 'content': 'Hello!'}],
)
print(response.message.content)

import ollama from 'ollama'

const response = await ollama.chat({
  model: 'qwen3-coder-next',
  messages: [{role: 'user', content: 'Hello!'}],
})
console.log(response.message.content)

Applications

Claude Code ollama launch claude --model qwen3-coder-next

OpenClaw ollama launch openclaw --model qwen3-coder-next

Hermes Agent ollama launch hermes --model qwen3-coder-next

Codex ollama launch codex --model qwen3-coder-next

OpenCode ollama launch opencode --model qwen3-coder-next

Models

View all →

Name

4 models

Size / Usage

Context

Input

qwen3-coder-next:latest

52GB · 256K context window · Text · 3 months ago

qwen3-coder-next:latest

52GB

256K

Text

qwen3-coder-next:cloud

High Usage · 256K context window · Text · 3 months ago

qwen3-coder-next:cloud

256K

Text

qwen3-coder-next:q4_K_M

52GB · 256K context window · Text · 3 months ago

qwen3-coder-next:q4_K_M

52GB

256K

Text

qwen3-coder-next:q8_0

85GB · 256K context window · Text · 3 months ago

qwen3-coder-next:q8_0

85GB

256K

Text

Readme

Built on top of Qwen3-Next-80B-A3B-Base, which adopts a novel architecture with hybrid attention and MoE, Qwen3-Coder-Next has been agentically trained at scale on large-scale executable task synthesis, environment interaction, and reinforcement learning, obtaining strong coding and agentic capabilities with significantly lower inference costs.

Features

Ultra-efficient inference: 80B total parameters, 3B active per token. Runs on consumer hardware with quantization.
256K native context: Full repository-scale understanding without chunking or retrieval hacks.
Agentic training: Trained on 800K executable tasks with environment interaction and reinforcement learning—not just static code-text pairs.
Tool calling: Works with coding agents like Claude Code, Qwen Code, Cline, and OpenCode out of the box.
Non-thinking mode only: Fast responses without <think></think> blocks.

Benchmarks

![image.png](/assets/library/qwen3-coder-next/7d9b5395-c335-48b9-b8c5-7e5090abbb8e)

### Features

* **Ultra-efficient inference:** 80B total parameters, 3B active per token. Runs on consumer hardware with quantization.
* **256K native context:** Full repository-scale understanding without chunking or retrieval hacks.
* **Agentic training:** Trained on 800K executable tasks with environment interaction and reinforcement learning—not just static code-text pairs.
* **Tool calling:** Works with coding agents like Claude Code, Qwen Code, Cline, and OpenCode out of the box.
* **Non-thinking mode only:** Fast responses without `<think></think>` blocks.

### Benchmarks

![](/assets/library/qwen3-coder-next/d978a788-2fa4-44ea-832c-20cd82202a6c)

Paste, drop or click to upload images (.png, .jpeg, .jpg, .svg, .gif)