GLM-4.6-FP8-Block

Creation

Made using LLM Compressor model_free_ptq:

from llmcompressor import model_free_ptq

MODEL_ID = "zai-org/GLM-4.6"
SAVE_DIR = MODEL_ID.rstrip("/").split("/")[-1] + "-FP8-BLOCK"

# Apply FP8-Block to the model
# Once quantized, the model is saved
# using compressed-tensors to the SAVE_DIR.
model_free_ptq(
    model_stub=MODEL_ID,
    save_directory=SAVE_DIR,
    scheme="FP8_BLOCK",
    ignore=[
        "re:.*gate$",
        "lm_head",
        "model.embed_tokens",
    ],
    max_workers=15,
    device="cuda:0",
)

Eval

GSM8k running with vLLM TP=4 on GB300

vllm serve GLM-4.6-FP8-BLOCK -tp=4

python tests/evals/gsm8k/gsm8k_eval.py     
Running GSM8K evaluation: 1319 questions, 5-shot
Evaluating: 100%|█████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 1319/1319 [00:52<00:00, 25.15it/s]

Results:
Accuracy: 0.929
Invalid responses: 0.001
Total latency: 52.449 s
Questions per second: 25.148
Total output tokens: 131828
Output tokens per second: 2513.428

GLM-4.6

👋 Join our Discord community.
📖 Check out the GLM-4.6 technical blog, technical report(GLM-4.5), and Zhipu AI technical documentation.
📍 Use GLM-4.6 API services on Z.ai API Platform.
👉 One click to GLM-4.6.

Model Introduction

Compared with GLM-4.5, GLM-4.6 brings several key improvements:

• Longer context window: The context window has been expanded from 128K to 200K tokens, enabling the model to handle more complex agentic tasks.
• Superior coding performance: The model achieves higher scores on code benchmarks and demonstrates better real-world performance in applications such as Claude Code、Cline、Roo Code and Kilo Code, including improvements in generating visually polished front-end pages.
• Advanced reasoning: GLM-4.6 shows a clear improvement in reasoning performance and supports tool use during inference, leading to stronger overall capability.
• More capable agents: GLM-4.6 exhibits stronger performance in tool using and search-based agents, and integrates more effectively within agent frameworks.
• Refined writing: Better aligns with human preferences in style and readability, and performs more naturally in role-playing scenarios.

We evaluated GLM-4.6 across eight public benchmarks covering agents, reasoning, and coding. Results show clear gains over GLM-4.5, with GLM-4.6 also holding competitive advantages over leading domestic and international models such as DeepSeek-V3.1-Terminus and Claude Sonnet 4.

Inference

Both GLM-4.5 and GLM-4.6 use the same inference method.

you can check our github for more detail.

Recommended Evaluation Parameters

For general evaluations, we recommend using a sampling temperature of 1.0.

For code-related evaluation tasks (such as LCB), it is further recommended to set:

• top_p = 0.95
• top_k = 40

Evaluation

• For tool-integrated reasoning, please refer to this doc.
• For search benchmark, we design a specific format for searching toolcall in thinking mode to support search agent, please refer to this. for the detailed template.