qwen3-next

316.6K Downloads Updated 2 months ago

The first installment in the Qwen3-Next series with strong performance in terms of both parameter efficiency and inference speed.

tools thinking cloud 80b

ollama run qwen3-next

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

from ollama import chat

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

import ollama from 'ollama'

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

Applications

Claude Code ollama launch claude --model qwen3-next

Codex ollama launch codex --model qwen3-next

OpenCode ollama launch opencode --model qwen3-next

OpenClaw ollama launch openclaw --model qwen3-next

Models

View all →

Name

10 models

Size

Context

Input

qwen3-next:latest

50GB · 256K context window · Text · 2 months ago

qwen3-next:latest

50GB

256K

Text

qwen3-next:80b

latest

50GB · 256K context window · Text · 2 months ago

qwen3-next:80b latest

50GB

256K

Text

qwen3-next:80b-cloud

256K context window · Text · 2 months ago

qwen3-next:80b-cloud

256K

Text

Readme

Qwen3-Next-80B-A3B is the first installment in the Qwen3-Next series and features the following key enhancements:

Hybrid Attention: Replaces standard attention with the combination of Gated DeltaNet and Gated Attention, enabling efficient context modeling for ultra-long context length.
High-Sparsity Mixture-of-Experts (MoE): Achieves an extreme low activation ratio in MoE layers, drastically reducing FLOPs per token while preserving model capacity.
Stability Optimizations: Includes techniques such as zero-centered and weight-decayed layernorm, and other stabilizing enhancements for robust pre-training and post-training.
Multi-Token Prediction (MTP): Boosts pretraining model performance and accelerates inference.