GLM-4.7 Flash 30B Uncensored — 16 GB VRAM, IQ3_M

An uncensored text-only build of GLM-4.7 Flash 30B, prepared for local use with Ollama.

This model uses IQ3_M quantization and is optimized for systems with approximately 16 GB of VRAM. It is intended for general text generation, coding, roleplay, analysis, experimentation and tool-calling workflows with fewer behavioral restrictions than the standard instruction-tuned release.

Model

byczech/glm-4.7-flash-uncensored-16G-AU-IQ3_M:30b

Highlights

• GLM-4.7 Flash 30B
• Text-only model
• Uncensored fine-tune
• Tool-calling support
• IQ3_M quantization
• Optimized for approximately 16 GB of VRAM
• Suitable for coding, agents, automation, roleplay, analysis and general chat

Run

ollama run byczech/glm-4.7-flash-uncensored-16G-AU-IQ3_M:30b

Recommended Ollama configuration

Flash Attention and a quantized KV cache can significantly reduce memory usage and allow a larger practical context window:

OLLAMA_FLASH_ATTENTION=1 \
OLLAMA_KV_CACHE_TYPE=q4_0 \
OLLAMA_NUM_PARALLEL=1 \
ollama serve

Then run the model in another terminal:

ollama run byczech/glm-4.7-flash-uncensored-16G-AU-IQ3_M:30b

You can verify whether the model is fully loaded on the GPU with:

ollama ps

The PROCESSOR column should ideally report 100% GPU.

Recommended settings for tool calling

For tool-calling and agentic workflows, the following sampling parameters are recommended:

temperature 0.7
top_p 1.0
min_p 0.01
repeat_penalty 1.0

These settings provide more consistent and reliable tool selection and argument generation than the default general-purpose configuration.

The model defaults are tuned for broader conversational and text-generation use:

temperature 1.0
top_p 0.95
min_p 0.01
repeat_penalty 1.0

Use the lower temperature and top_p 1.0 configuration when predictable tool calls are more important than response diversity.

API example

curl http://localhost:11434/api/chat -d '{
  "model": "byczech/glm-4.7-flash-uncensored-16G-AU-IQ3_M:30b",
  "messages": [
    {
      "role": "user",
      "content": "Analyze this task and choose the appropriate tool."
    }
  ],
  "stream": false
}'

Tool-calling example with Python

Install the Ollama Python package:

pip install ollama

from ollama import chat


def get_weather(city: str) -> str:
    """Return a mock weather report for a city."""
    return f"The weather in {city} is sunny."


response = chat(
    model="byczech/glm-4.7-flash-uncensored-16G-AU-IQ3_M:30b",
    messages=[
        {
            "role": "user",
            "content": "What is the weather in Prague?",
        }
    ],
    tools=[get_weather],
    options={
        "temperature": 0.7,
        "top_p": 1.0,
        "min_p": 0.01,
        "repeat_penalty": 1.0,
    },
)

print(response.message)

Tool execution is handled by the client application. The model selects a tool and generates its arguments, while your application remains responsible for executing the function and returning the result.

Recommended settings for tool calling

For tool-calling and agentic workflows, the following sampling parameters are recommended:

temperature 0.7
top_p 1.0
min_p 0.01
repeat_penalty 1.0

These settings provide more consistent and reliable tool selection and argument generation than the default general-purpose configuration.

The model defaults are tuned for broader conversational and text-generation use:

temperature 1.0
top_p 0.95
min_p 0.01
repeat_penalty 1.0

Use the lower temperature and top_p 1.0 configuration when predictable tool calls are more important than response diversity.

Uncensored model notice

This build is designed to apply fewer response restrictions than the standard model. As a result, it may generate inaccurate, offensive, explicit or otherwise undesirable content more readily.

Use it responsibly and apply your own validation, moderation and access controls when integrating it into applications, especially when tool execution or external actions are enabled.

Quality and memory trade-off

The IQ3_M quantization was selected to balance model quality and VRAM usage while keeping the model practical on a 16 GB GPU.

Using a Q4_0 KV cache further reduces memory consumption and leaves more VRAM available for context. For higher KV-cache precision, use q8_0, but expect higher VRAM usage and a smaller practical context window.

Notes

• Actual VRAM usage depends on the Ollama version, GPU, backend, driver and runtime configuration.
• Keep OLLAMA_NUM_PARALLEL=1 for the lowest memory usage.
• Close other GPU-intensive applications when maximizing context length.
• Tool calls should always be validated before execution.
• Uncensored does not mean more accurate or more capable; it primarily means fewer behavioral restrictions.