# Qwen3.6

> Alibaba's open-weight coding line — a 27B dense model with flagship-level agentic coding

- Maker: Alibaba (Qwen)
- Line: Qwen (open-weight)
- Canonical page: https://ai-tldr.dev/models/qwen3-6/
- Updated: 2026-06-18

## In short

Qwen3.6 is the April 2026 release from the Qwen team at Alibaba Group, building on Qwen3.5. It is an open-weight line published under the Apache 2.0 license, distributed on Hugging Face and ModelScope and served through Alibaba Cloud Model Studio. It costs $0.2885 per million input tokens and $3.17 per million output tokens. You can use it through Alibaba Cloud Model Studio, OpenRouter and Hugging Face (self-host).

## Facts

| Fact | Value | Source |
| --- | --- | --- |
| Context window | 262K (extensible ~1M) (max output ~32K (up to ~82K for complex reasoning)) | [1] |
| Input price | $0.2885 / 1M tokens | [2] |
| Output price | $3.17 / 1M tokens | [2] |
| Licence | Apache 2.0 · open weights | [1] |
| Released | April 2026 | [3] |
| Parameters | 27B dense (Qwen3.6-27B); 35B total / 3B active MoE (Qwen3.6-35B-A3B) | [1] |
| Knowledge cutoff | Not publicly disclosed | [1] |

## Sources

1. [Qwen3.6-27B model card (Hugging Face)](https://huggingface.co/Qwen/Qwen3.6-27B)
2. [Qwen3.6-27B on OpenRouter](https://openrouter.ai/qwen/qwen3.6-27b)
3. [Qwen3.6-27B release blog](https://qwen.ai/blog?id=qwen3.6-27b)

## Overview

Qwen3.6 is the April 2026 release from the Qwen team at Alibaba Group, building on Qwen3.5. It is an open-weight line published under the Apache 2.0 license, distributed on Hugging Face and ModelScope and served through Alibaba Cloud Model Studio. The release leads with two open models: Qwen3.6-35B-A3B, a Mixture-of-Experts model with 35B total and roughly 3B active parameters (released 2026-04-16), and Qwen3.6-27B, a 27-billion-parameter dense model (released 2026-04-22) that the Qwen team frames as 'flagship-level coding in a 27B dense model.'

The headline story is parameter efficiency in agentic coding: the dense Qwen3.6-27B reaches 77.2 on SWE-bench Verified and 59.3 on Terminal-Bench 2.0, edging past the much larger Qwen3.5-397B-A17B MoE from the previous generation on most coding benchmarks. Qwen3.6 focuses on real-world utility — smoother front-end workflows, repository-level reasoning, and a 'thinking preservation' feature that keeps reasoning context across conversation history for iterative, agentic development.

Architecturally, Qwen3.6 uses a hybrid attention stack (Gated DeltaNet linear attention interleaved with periodic full Gated Attention) and Multi-Token Prediction for faster decoding. Both released models carry a vision encoder, accept text, image, and video input, and serve a native 262,144-token context that can be stretched toward ~1M tokens with YaRN RoPE scaling. As an Apache-2.0 release, the weights can be self-hosted via vLLM, SGLang, llama.cpp, MLX, or Transformers, or accessed through hosted APIs.

## Benchmarks

- SWE-bench Verified: 77.2% ([source](https://qwen.ai/blog?id=qwen3.6-27b))
- Terminal-Bench 2.0: 59.3% ([source](https://qwen.ai/blog?id=qwen3.6-27b))
- MMLU-Pro: 86.2% ([source](https://huggingface.co/Qwen/Qwen3.6-27B))
- AIME 2026: 94.1% ([source](https://huggingface.co/Qwen/Qwen3.6-27B))
- GPQA Diamond: 87.8% ([source](https://huggingface.co/Qwen/Qwen3.6-27B))
- MMMU Pro: 75.8% ([source](https://huggingface.co/Qwen/Qwen3.6-27B))

## Pricing

- Input: $0.2885 / 1M tokens per 1M tokens
- Output: $3.17 / 1M tokens per 1M tokens

Qwen3.6-27B via OpenRouter (lowest listed provider); prices vary by host. The Apache-2.0 weights can also be self-hosted at no per-token cost.

Source: https://openrouter.ai/qwen/qwen3.6-27b

## Strengths

- Flagship-level agentic coding at small scale — the 27B dense model scores 77.2 on SWE-bench Verified, beating the prior-gen 397B MoE on most coding benchmarks
- Apache 2.0 open weights — free to self-host, fine-tune, and deploy commercially with no license gate
- Long context out of the box (262K native, extensible toward ~1M via YaRN scaling)
- Multimodal input — text, images, and video accepted via a built-in vision encoder
- Efficient hybrid architecture (Gated DeltaNet + periodic full attention) with Multi-Token Prediction for lower-latency decoding
- Thinking-preservation keeps reasoning context across turns, suited to iterative agentic workflows
- Broad ecosystem support: vLLM, SGLang, llama.cpp, MLX, Transformers, plus Qwen Code and Qwen-Agent

## Best for

- Agentic coding and repository-level software engineering (SWE-bench / Terminal-Bench style tasks)
- Front-end and full-stack web development workflows
- Self-hosted, privacy-sensitive deployments where Apache-2.0 weights matter
- Long-document and large-codebase analysis using the 262K-token context
- Multimodal tasks combining code, images, and video understanding
- Fine-tuning a strong open base for domain-specific coding or reasoning agents
- Cost-efficient inference where a 27B dense or 3B-active MoE beats running a much larger model

## How to access

- [Alibaba Cloud Model Studio](https://modelstudio.alibabacloud.com/) — model id `qwen3.6-27b`
- [OpenRouter](https://openrouter.ai/qwen/qwen3.6-27b) — model id `qwen/qwen3.6-27b`
- [Hugging Face (self-host)](https://huggingface.co/Qwen/Qwen3.6-27B) — model id `Qwen/Qwen3.6-27B`

## Qwen (open-weight) versions

- Qwen3.6 (April 2026) — this page
- [Qwen3.5](https://ai-tldr.dev/models/qwen3-5/) (16 February 2026)
- [Qwen3 (2507 update)](https://ai-tldr.dev/models/qwen3-2507/) (July 2025)
- [Qwen3](https://ai-tldr.dev/models/qwen3/) (28 April 2025)
- [Qwen2.5](https://ai-tldr.dev/models/qwen2-5/) (September 2024)
- [Qwen2](https://ai-tldr.dev/models/qwen2/) (June 2024)

## FAQ

### Is Qwen3.6 open source?

The released Qwen3.6 models — Qwen3.6-27B (dense) and Qwen3.6-35B-A3B (MoE) — are open-weight under the Apache 2.0 license. You can download them from Hugging Face or ModelScope and self-host, fine-tune, or deploy commercially. (Note: Alibaba's separate Qwen3.6-Max flagship is a closed-weight, API-only model.)

### What is the difference between Qwen3.6-27B and Qwen3.6-35B-A3B?

Qwen3.6-27B is a 27-billion-parameter dense model and the team's 'flagship-level' coding model. Qwen3.6-35B-A3B is a Mixture-of-Experts model with 35B total parameters but only ~3B active per token, trading a bit of peak accuracy for cheaper, faster inference. Both share the hybrid attention architecture, vision input, and 262K context.

### How long is Qwen3.6's context window?

Both open models serve a native context of 262,144 tokens and can be extended toward roughly 1 million tokens using YaRN RoPE scaling. Qwen recommends keeping at least 128K of context to preserve the model's reasoning behavior.

### How good is Qwen3.6 at coding?

The dense Qwen3.6-27B scores 77.2 on SWE-bench Verified and 59.3 on Terminal-Bench 2.0, beating the prior-generation Qwen3.5-397B-A17B MoE on most coding benchmarks despite being far smaller. It is tuned for agentic coding, front-end workflows, and repository-level reasoning.

## Links

- [Qwen3.6-27B release blog](https://qwen.ai/blog?id=qwen3.6-27b)
- [Qwen3.6-35B-A3B release blog](https://qwen.ai/blog?id=qwen3.6-35b-a3b)
- [Qwen3.6 on GitHub](https://github.com/QwenLM/Qwen3.6)
- [Qwen3.6-27B model card (Hugging Face)](https://huggingface.co/Qwen/Qwen3.6-27B)
- [Qwen3.6-35B-A3B model card (Hugging Face)](https://huggingface.co/Qwen/Qwen3.6-35B-A3B)
- [Alibaba Cloud Model Studio (API)](https://modelstudio.alibabacloud.com/)
- [Qwen3.6-27B on OpenRouter](https://openrouter.ai/qwen/qwen3.6-27b)
- [Qwen Studio (chat)](https://chat.qwen.ai/)
- [the-decoder: Qwen3.6-27B beats larger predecessor](https://the-decoder.com/qwen3-6-27b-beats-much-larger-predecessor-on-most-coding-benchmarks/)

Source: AI/TLDR LLM registry, https://ai-tldr.dev/models/qwen3-6/