Qwen 3.6 27B

Overview

Dense 27B multimodal model from Alibaba (April 2026), scoring 77.2% on SWE-bench Verified with 262k native context (1M via YaRN). The Qwen 3.6 generation's developer-friendly workhorse.

When to pick this model

• Coding agents needing top-tier SWE-bench accuracy at single-GPU scale
• Multimodal applications with long context (up to 1M with YaRN)
• Apache 2.0 deployments replacing closed APIs
• Reasoning workloads where dense models behave more predictably than MoE
• Local inference on a single 24-32 GB GPU

VRAM requirements by quantization

Quantization	VRAM required
Q4_K_M (recommended)	16 GB
Q5_K_M	19 GB
Q8_0	29 GB
FP16 (no quantization)	54 GB

VRAM figures include model weights plus a typical 8k KV cache and ~600 MB runtime overhead (Ollama / llama.cpp baseline). Add headroom for higher context lengths.

Published benchmark scores

Benchmark	Score
SWE-bench Verified	77.2
Terminal-Bench	59.3
SkillsBench	48.2

Scores published by the model author or aggregated from public leaderboards. Re-measured monthly by our editorial team.

Strengths

• 77.2% SWE-bench Verified — frontier coding accuracy
• Native multimodal text + image
• 262k context, extendable to 1M with YaRN
• Apache 2.0
• Dense 27B fits comfortably on consumer hardware

Limitations

• Needs 16+ GB VRAM at Q4
• Hybrid architecture requires a recent llama.cpp build
• Dense design means no MoE inference efficiency

Architecture & training

Architecture: Dense 27B · Gated DeltaNet + Gated Attention · multimodal · 64 layers

Training: Dense successor to Qwen 3.5 27B, 3.6 generation.

Quick start

ollama run qwen3.6:27b

Or use the open-source MCP server to query this model from Claude Desktop, Cursor, or any MCP-compatible client.