NobodyWho/Qwen_Qwen3.6-27B-GGUF

Overview

GGUF quantization of Qwen3.6-27B, prepared for NobodyWho: it works with NobodyWho out of the box, with Qwen's recommended sampling metadata embedded in every quant, and is verified with NobodyWho's test suite. Qwen3.6-27B is Alibaba's dense flagship — natively multimodal (text + image), with strong reasoning and best-in-class agentic tool calling (it rivals far larger MoE models on agentic-coding benchmarks).

Model Capabilities

• Text generation — instruction-following chat
• Tool calling — native function calling with grammar-constrained output (14/14 on NobodyWho's suite)
• Vision — image understanding via the companion mmproj-BF16.gguf projection model
• Reasoning — thinking mode (on by default)
• Long context — up to 256k tokens
• Multilingual — broad language coverage

Available Quantizations

Verified with NobodyWho's suite — tool calling 14/14 on Q3_K_M and Q4_K_M, vision on Q3_K_M (June 2026); Q8_0 not separately run (RAM). Q3_K_M (≈13.6 GB) is the comfortable fit on 24 GB; Q4_K_M (≈16.8 GB) also runs on 24 GB but is tight — it swaps and runs slower (tested 14/14); Q8_0 (≈28.6 GB) wants 32 GB+. BF16 (≈54 GB) is not hosted. The upstream GGUF has no general.sampling.* metadata, so all quants embed Qwen's recommended sampler (see INJECTION.md).

Quick Start

Using the NobodyWho library:

from nobodywho import Chat

chat = Chat("huggingface:NobodyWho/Qwen_Qwen3.6-27B-GGUF/Qwen_Qwen3.6-27B-Q3_K_M-vendor-sampling.gguf")
response = chat.ask("What is the capital of Denmark?").completed()
print(response)  # The capital of Denmark is Copenhagen.

Vision

from nobodywho import Model, Chat, Prompt, Image, Text

model = Model(
    "huggingface:NobodyWho/Qwen_Qwen3.6-27B-GGUF/Qwen_Qwen3.6-27B-Q3_K_M-vendor-sampling.gguf",
    projection_model_path="huggingface:NobodyWho/Qwen_Qwen3.6-27B-GGUF/mmproj-BF16.gguf",
)
chat = Chat(model=model, system_prompt="You are a helpful assistant.")
response = chat.ask(Prompt([
    Text("What is in this image?"),
    Image("./photo.png"),
])).completed()
print(response)

llama-cpp-python

from llama_cpp import Llama

llm = Llama.from_pretrained(
    repo_id="NobodyWho/Qwen_Qwen3.6-27B-GGUF",
    filename="Qwen_Qwen3.6-27B-Q3_K_M-vendor-sampling.gguf",
)

Model Specifications

• Parameters: 27B (dense)
• Context length: 262,144 tokens (256K)
• License: Apache 2.0
• Base model: Qwen/Qwen3.6-27B
• Architecture: qwen35 (vision-capable)

Licensing / Credits

Licensed under Apache 2.0 (unchanged from upstream). All model credit belongs to the Qwen team, Alibaba Group. GGUF quantizations provided by unsloth.