GLM-5.1

Description:

GLM-5.1 is a next-generation flagship large language model designed for agentic engineering, featuring significantly stronger coding capabilities than its predecessor GLM-5. It sustains optimization over hundreds of reasoning rounds and thousands of tool calls, enabling it to solve complex software engineering tasks with extended reasoning horizons and iterative refinement.

GLM-5.1 was developed by Z.ai (zai-org) as a part of the GLM model family.

This model is ready for commercial/non-commercial use.

Third-Party Community Consideration

This model is not owned or developed by NVIDIA. This model has been developed and built to a third-party's requirements for this application and use case; see link to Non-NVIDIA GLM-5.1 Model Card.

License/Terms of Use:

GOVERNING TERMS: The trial service is governed by the NVIDIA API Trial Terms of Service; use of this model is governed by the NVIDIA Open Model License Agreement. ADDITIONAL INFORMATION: MIT License.

Deployment Geography:

Global

Use Case:

Developers and researchers can use GLM-5.1 for complex software engineering tasks, agentic workflows requiring long-horizon planning and tool use, mathematical reasoning, terminal-based automation, code generation and debugging, and general conversational AI applications.

Release Date:

Build.NVIDIA.com: 4/17/2026 via link
NGC: 04/15/2026 via link

HuggingFace: 04/08/2026 via link

Reference(s):

References:

Model Architecture:

Architecture Type: Mixture of Experts (MoE)

Network Architecture: GLM (General Language Model) with DSA (Dense-Sparse-Alternating)

This model was developed based on GLM-5.

Number of model parameters: 754B

Input:

Input Type(s): Text

Input Format(s): String

Input Parameters: One Dimensional (1D)

Other Properties Related to Input: Supports multi-turn conversations, tool calling, system prompts, and extended agentic sessions. Input context length: 131,072 tokens.

Output:

Output Type(s): Text

Output Format: String

Output Parameters: One Dimensional (1D)

Other Properties Related to Output: Supports streaming, structured output, reasoning traces, and tool call responses. Output context length: 131,072 tokens.

Our AI models are designed and/or optimized to run on NVIDIA GPU-accelerated systems. By leveraging NVIDIA's hardware (e.g. GPU cores) and software frameworks (e.g., CUDA libraries), the model achieves faster training and inference times compared to CPU-only solutions.

Software Integration:

Runtime Engine(s):

vLLM

Supported Hardware Microarchitecture Compatibility:

NVIDIA Blackwell
NVIDIA Hopper

Supported Operating System(s):

Linux

The integration of foundation and fine-tuned models into AI systems requires additional testing using use-case-specific data to ensure safe and effective deployment. Following the V-model methodology, iterative testing and validation at both unit and system levels are essential to mitigate risks, meet technical and functional requirements, and ensure compliance with safety and ethical standards before deployment.

Model Version(s):

GLM v5.1

Training, Testing, and Evaluation Datasets:

Training Dataset

Data Modality: Text (English, Chinese, multilingual)
Text Training Data Size: Undisclosed
Training Data Collection: Undisclosed
Training Labeling: Undisclosed
Training Properties: Undisclosed

Testing Dataset

Testing Data Collection: Undisclosed
Testing Labeling: Undisclosed
Testing Properties: Undisclosed

Evaluation Dataset

Evaluation Benchmark Score: Evaluated on multiple benchmarks including SWE-Bench Pro, Terminal-Bench 2.0, CyberGym, NL2Repo, AIME 2026, HMMT, GPQA-Diamond, IMOAnswerBench, HLE w/ Tools, BrowseComp, τ³-Bench, MCP-Atlas, and Tool-Decathlon.
Evaluation Data Collection: Automated
Evaluation Labeling: Human
Evaluation Properties: GLM-5.1 achieves state-of-the-art results across coding, mathematical reasoning, and agentic benchmarks, outperforming its predecessor GLM-5 on all major evaluations.

Evaluation Results

Detailed Benchmark Results

Coding & Engineering

Benchmark	Dataset	GLM-5.1	GLM-5
SWE-Bench Pro	ScaleAI/SWE-bench_Pro	58.4%	55.1%
NL2Repo	—	42.7%	35.9%
Terminal-Bench 2.0	harborframework/terminal-bench-2.0	63.5%	56.2%
Terminal-Bench 2.0 (Claude Code)	harborframework/terminal-bench-2.0	69.0%	—
CyberGym	—	68.7%	—

Math & Reasoning

Benchmark	Dataset	GLM-5.1
AIME 2026	MathArena/aime_2026	95.3%
HMMT Nov. 2025	—	94.0%
HMMT Feb. 2026	MathArena/hmmt_feb_2026	82.6%
GPQA-Diamond	Idavidrein/gpqa	86.2%
IMOAnswerBench	—	83.8%

Tool Use & Agentic

Benchmark	Dataset	GLM-5.1
HLE	cais/hle	31.0%
HLE (w/ Tools)	cais/hle	52.3%
BrowseComp	—	68.0%
BrowseComp (w/ Context Manage)	—	79.3%
Tool-Decathlon	—	40.7%
τ³-Bench	—	70.6%
MCP-Atlas (Public Set)	—	71.8%

Inference:

Acceleration Engine: vLLM

Test Hardware:

NVIDIA GB200x4

Ethical Considerations:

NVIDIA believes Trustworthy AI is a shared responsibility and we have established policies and practices to enable development for a wide array of AI applications. When downloaded or used in accordance with our terms of service, developers should work with their internal model team to ensure this model meets requirements for the relevant industry and use case and addresses unforeseen product misuse.

Please report model quality, risk, security vulnerabilities or NVIDIA AI Concerns here.