Skip to content
Alibaba (Qwen)

Qwen 3.6 27B

Open Source

Alibaba (Qwen)•Released on 2026-04-22

Alibaba's open-weight 27B dense flagship released April 22, 2026. Delivers flagship-level agentic coding in a single dense checkpoint with hybrid thinking, 262K native context (extensible to 1M), and a vision encoder. Released under Apache-2.0 on HuggingFace.

Core Specs

262K
Context Window
33K
Max Output
ReasoningOpen Sourcetextimage

Scenario Scores

Coding
—
Writing
—
Translation
—
Data Analysis
—
Conversation
—
Image Gen
—

Pros & Cons

Sentiment0% +0% ·0% −

Pros

  • +Apache-2.0 fully open weights
  • +Outperforms Qwen3.5-397B-A17B on coding
  • +262K native context (1M extended)
  • +Dense architecture — no MoE routing complexity
  • +Runs on consumer GPUs via GGUF quantization

Cons

  • −Released <24h ago — real-world feedback thin
  • −No hosted API pricing yet (self-host / community providers only)
  • −27B still needs ~18GB VRAM even quantized

Pricing

Input (per 1M tokens)$0.00
Output (per 1M tokens)$0.00
Free trial available
Updated on 2026-04-23

Get Started

1Visit the provider's website
2Create an account
3Start using the model

Benchmarks

sweBenchVerified77.2%
terminalBench59.3%
skillsBench48.2%

Reliability

Incidents (30d)0

Compare with Others

Qwen 3.6 27B vs Claude Opus 4.7 →Qwen 3.6 27B vs Claude Opus 4.6 →Qwen 3.6 27B vs Claude Opus 4.5 →