Model Detail Human reviewed

Qwen3.6 27B

Alibaba's open-weight Qwen3.6 27B. Strong coding-agent, reasoning, long-context, and vision-language model with 262K native context.

AlibabaQWEN3apache-2.02026-04
Parameters
27B
Dense model
Context window
262K
Long context
Architecture
hybrid linear attention transformer with vision encoder
dense
Quality score
94
Planner signal

Task Fit

Code AgentSupported

Tool use, repo work, terminal workflows, and coding benchmarks.

CodeSupported

Code generation, debugging, refactoring, and benchmark signal.

ChatSupported

General writing, Q&A, and assistant use.

RAGSupported

Document QA benefits from long context and instruction following.

VisionSupported

Image or visual understanding, not necessarily image generation.

Image GenerationNot a fit

Not marked for image generation in the current library.

Video GenerationNot a fit

Not marked for video generation in the current library.

VoiceNot a fit

Not marked for voice in the current library.

Source Confidence

Overallhigh · 100/100
ParametersReviewed / seeded
Task fitReviewed / seeded
MemorySeeded artifact
LicenseSource / seed
BenchmarksAvailable
Hardware fitCalculated

Variants and Quant Artifacts

Choose the artifact first; hardware fit follows from RAM, VRAM, format, and runtime.

4 artifacts
QuantFormatQualityMin RAMReco RAMRuntimeAction
Q4_K_Mggufbalanced24GB32GBllama.cpp, lm-studio Plan with this
Q5_K_Mggufbalanced26GB40GBllama.cpp, lm-studio Plan with this
Q8_0ggufhigh36GB48GBllama.cpp, lm-studio Plan with this
FP16safetensorshigh64GB96GBtransformers, vllm, sglang Plan with this

Recommended Hardware

Cheapest That Works
NVIDIA RTX 4060 Ti 16GB
Compatible

Lowest estimated 5-year cost that can run this model.

16GB VRAM$1,772 / 5y
Best Value
NVIDIA Jetson AGX Orin 64GB
Compatible

Enough unified/system memory with a balanced 5-year cost.

64GB RAM$2,553 / 5y
Best Performance
NVIDIA RTX 6000 Blackwell 96GB
Compatible

Highest local performance signal among compatible hardware.

96GB VRAM$12,376 / 5y

Benchmarks

Aime2694.1%
GPQA Diamond87.8%
MMLU Pro86.2%
Livecodebench83.9%
Mmmu82.9%
Swe Bench77.2%
Androidworld70.3%
Terminal Bench59.3%

Source and Review

Hugging FaceQwen/Qwen3.6-27B
OllamaNot mapped
VerificationHuman reviewed
Artifact sourcehuman-estimated-from-hf-card
Default variantQwen3.6 27B Coder
Tool callingSupported

Similar Models