Model

GLM-4

CLOSED

Zhipu AI / THUDM (Tsinghua) · GLM-4 family · released Jan 16, 2024

Original closed flagship unveiled at Zhipu DevDay, 128K context + 'All Tools' agent, positioned near GPT-4; since retired from the API in favor of GLM-4-Long/Flash and the open GLM-4.x/GLM-5 lines.

ReasoningCodingVisionFunction callingTool useAgentic

976.2

Elo · rank #218

Parameters

Undisclosed

Active params

—

Context

128K tokens

Architecture

Dense Transformer (undisclosed size; architecture details not published)

License

Proprietary (Zhipu AI commercial API)

Languages

—

API price (in/out)

No hosted API

Modalities

text

Benchmark results

Bar shows position within the tracked field; marker = field best

BIG-Bench HardReasoning84.7%#18

best: ERNIE 4.5 300B-A47B · 94.3%

BFCLAgents81.8%#9

best: SmolLM3 3B (No Thinking) · 92.3%

GPQA DiamondReasoning39.9%#220

best: GPT-5.6 · 94.6%

GSM8KMath93.3%#21

best: Llama 3.1 405B · 96.8%

HumanEvalCoding78.5%#58

best: Claude Opus 4.5 · 99.4%

IFEvalReasoning83.7%#73

best: Gemma 4 26B A4B · 98.5%

MATH-500Math61.3%#109

best: GPT-5 · 99.4%

MMLUKnowledge83.3%#70

best: OpenAI o3 · 92.9%

Run it locally

Closed weights — available via API only. No local deployment.

Input / M tok

—

Output / M tok

—

GLM-4 family

Elo progression across releases

GLM-4Jan 2024976.2

Resources

DOCSdocs.bigmodel.cn

API price — · each benchmark row carries its own source badge (see methodology)