Qwen3.6 27B A4B is a dense 27B-parameter model from the Qwen Team, aimed at agentic coding, reasoning, and multimodal tasks. It accepts text, image, and video input, includes an optional thinking mode for harder problems, and supports a 159K context window across many languages.
Qwen3.6 27B A4B - NVIDIA RTX 5090
qwen/qwen3.6-27b
qwen3.6:27bReleased Apr 27, 2026159K context
Best forCodeAgentic Coding
Best fit competitor·Claude 3.5 Sonnet
Why teams pick Qwen3.6 27B A4B over Claude 3.5 Sonnet
Built for agentic coding and repo-level reasoning — the same workloads teams reach for Sonnet on, without sending code to a third-party API.
262K native context (YaRN to ~1M) on hardware you control, with Thinking Preservation across long sessions.
Apache 2.0 and self-hosted: fine-tune, air-gap, and ship features without Anthropic rate limits or data-handling policies.
About
Qwen3.6 27B A4B is a dense 27-billion-parameter language model from the Qwen Team at Alibaba, released in April 2026. It features hybrid multimodal capabilities — accepting text, image, and video inputs — and supports a 262,144-token context window.
The model is designed for agentic coding and reasoning tasks, with particular strength in repository-level code comprehension, front-end development workflows, and multi-step problem solving. It includes a built-in thinking mode for extended reasoning and preserves thinking context across conversation history. Qwen3.6 27B A4B supports 201 languages and dialects and is released under the Apache 2.0 license.
Compare
Model Cost Across Durations
Live pack pricing vs typical API estimates from 11 hours through 1 month.
API estimates for GPT-5.4 and Claude Opus 4.6 vs Qwen 3.6 27B A4B on RTX 5090.
Time pack
Qwen 3.6 27B A4B on RTX 5090
24 hours cost
$31
Lowest
GPT-5.4
24 hours cost
$35.47
Save $4.47 vs our model
Claude Opus 4.6
24 hours cost
$63.07
Save $32.07 vs our model
Models in chart
- Qwen 3.6 27B A4B on RTX 5090
- GPT-5.4
- Claude Opus 4.6
At a glance
Release
Apr 27, 2026
Parameters
27.8 B (reported)
Quantization
Q4_K_M
Size
17GB
Context
159K
Benchmarks
Snapshot of third-party and official benchmark metrics for Qwen3.6 27B A4B.
Performance indexes
46
Artificial Analysis
Intelligence Index
#1 among open-weight small models (4B–40B)
77.2
SWE-bench Verified
Coding Index
Strong real-world GitHub issue resolution
59.3
Terminal-Bench 2.0
Agentic Index
Agentic terminal + tool use
Benchmark scores
GPQA Diamond
iGraduate-level scientific reasoning
87.8%
HLE
iHumanity's Last Exam
24.0%
IFBench
iInstruction-following benchmark
83.9%
What it's good at
Flagship-level agentic coding and terminal use — outperforms the 397B Qwen3.5 MoE on every major coding benchmark.
Repository-level code comprehension and multi-step problem solving across long contexts.
Front-end development: QwenWebBench covers Web Design, Web Apps, Games, SVG, Data Visualization, Animation, and 3D (bilingual EN/CN).
Extended reasoning via built-in Thinking Preservation mode — preserves chain-of-thought across conversation turns.
Strong multilingual support: 201 languages and dialects.
Efficient local deployment: runs on ~18GB VRAM (Q4 quantized: ~16.8GB); dense architecture compresses more predictably than MoE.
Apps & integrations
Choose an app below. Each guide shows how to point the app at your OpenAI-compatible endpoint.
n8n
Automate workflows and call your model as a node.
Open
OpenClaw
Build AI agents and tools on an OpenAI-compatible endpoint.
Open
Hermes
Connect agent runners to your chat completions endpoint.
Open
OpenCode
Power developer tools with your OpenAI-compatible model.
Open
Cursor
Override OpenAI Base URL in Cursor Settings and use your model with BYOK.
Open
VS Code
Use the Cline extension in VS Code to connect your OpenAI-compatible endpoint.
Open
Codex
Run OpenAI Codex CLI against your Chat Completions endpoint via config.toml.
Open
Raspberry Pi
Full Pi OS guide: SSH, API keys, curl, Python venv, systemd, and troubleshooting.
Open
FAQ
Frequently asked questions
Common questions about Qwen3.6 27B A4B, deployment, and using it on OpenLLM Buddy.
6 questions
Use the Deploy button on this page or go to the console, select a pack, and start an instance. Once the deployment is live, copy your chat-completions endpoint and API key from the console to call the model from your app or workflow tools.
Qwen3.6 27B A4B is released under . Running it on OpenLLM Buddy gives you a managed GPU instance; refer to Alibaba’s model documentation for upstream terms and attribution.
Point any OpenAI-compatible client at your instance’s /v1/chat/completions URL. Use qwen3.6:27b as the model name in requests—the identifier listed on this page. The API tab includes curl and SDK examples you can copy.
Billing is flat per deployment pack (time-based), not pay-per-token like hosted APIs. Open the Pricing tab for pack durations and prices. Usage counters help you monitor load but do not change the pack price you selected at deploy.
Yes. Qwen3.6 27B A4B is tuned for repository-level code understanding, multi-step reasoning, and front-end workflows. Pair it with the OpenAI-compatible API and your agent framework of choice, or connect via the Apps & integrations section on the Overview tab.
Ready to try it? Deploy Qwen3.6 27B A4B · Browse models