Model

GLM-5.1

OPEN

Zhipu AI / Z.ai (Tsinghua) · GLM-5 family · released Apr 7, 2026

Point release focused on extreme long-horizon agentic engineering; Z.ai's own SWE-bench Pro claim (58.4) briefly topped the public leaderboard, edging out GPT-5.4 and Claude Opus 4.6 on that benchmark at launch.

ReasoningCodingVisionFunction callingTool useAgentic

2241.1

Elo · rank #40

Parameters

744B

Active params

40B (MoE)

Context

200K tokens

Architecture

Mixture-of-Experts, 744B total / 40B active, DeepSeek Sparse Attention (DSA), 200K context, tuned for autonomous long-horizon agent runs (up to 8 hours single-run)

License

MIT

Languages

—

API price (in/out)

$1.4 / $4.4

Modalities

text

Benchmark results

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

Agents' Last ExamAgents28.1%#13

best: GPT-5.6 · 52.7%

AIMEMath95.3%#13

best: GPT-5.2 · 100.0%

BrowseCompAgents68.0%#23

best: Kimi K3 · 91.2%

CyberGymCoding68.7%#3

best: GPT-5.6 · 84.5%

GDPval-AAAgents1260#25

best: Claude Fable 5 · 1932

GPQA DiamondReasoning86.2%#51

best: GPT-5.6 · 94.6%

Humanity's Last ExamReasoning31.0%#44

best: Claude Sonnet 5 · 57.4%

IFBenchReasoning76.0%#13

best: MiniMax M3 · 83.0%

LiveCodeBenchCoding81.4%#44

best: DeepSeek-V4-Pro (Think Max) · 93.5%

SWE-bench ProCoding58.4%#13

best: Claude Fable 5 · 80.0%

SWE-bench VerifiedCoding76.4%#36

best: Claude Fable 5 · 95.0%

Terminal-Bench 2.0Coding63.5%#27

best: GPT-5.6 · 88.8%

Run it locally

VRAM @ Q4

—

VRAM @ FP16

—

Fits on (Q4)

Multi-node cluster required

Throughput data unavailable.

Quantizations

—

GLM-5 family

Elo progression across releases

GLM-5Feb 20262123.1 GLM-5.1Apr 20262241.1 GLM-5.2 (High)Jun 20261000.0 GLM-5.2 (Max)Jun 20262455.0

Resources

HFzai-org/GLM-5.1 DOCShttps://docs.z.ai/guides/llm/glm-5.1 APIAvailable via Zhipu AI / Z.ai (Tsinghua) API

API price $1.4/$4.4 · each benchmark row carries its own source badge (see methodology)