Z.ai releases GLM-5, an MIT-licensed open model for…

Z.ai (Zhipu) GLM-5 at a glance

GLM-5 scales to 744B total and 40B active parameters, adds DeepSeek Sparse Attention, and ships under the MIT license on Hugging Face.

What changed with Z.ai (Zhipu) GLM-5

Z.ai's release notes list GLM-5 on February 12, 2026. The model card describes 744B total and 40B active parameters, up from GLM-4.5's 355B, with DeepSeek Sparse Attention for token efficiency and a focus on complex systems engineering and long-running agent tasks.

Z.ai reports results such as 77.8 on SWE-bench Verified and 56.2 on Terminal-Bench 2.0, and claims the best open-source scores on reasoning, coding, and agent tests at release. These are self-reported. Weights are on Hugging Face under the MIT license.

Z.ai (Zhipu) GLM-5 benchmarks and pricing

Values are copied from the linked source on the date shown. They are not IA Compa measurements.

Output speedas of Nov 5, 2026 · Source for Output speed: artificialanalysis.ai (opens in a new tab): 62 tokens/s
Input price per 1M tokensas of Nov 5, 2026 · Source for Input price per 1M tokens: artificialanalysis.ai (opens in a new tab): $1.00
Output price per 1M tokensas of Nov 5, 2026 · Source for Output price per 1M tokens: artificialanalysis.ai (opens in a new tab): $3.20

How it compares

Other models in the same category that we have covered, using each model's latest recorded value. (Models)

Output speed (tokens/s)higher is better

Claude Haiku 5.5236 tokens/s

Claude Sonnet 5.5153 tokens/s

Muse Wonder123 tokens/s

Gemini 3.8 Flash115 tokens/s

Mistral Large 4113 tokens/s

GPT-6 Sol111 tokens/s

Claude Opus 5.5106 tokens/s

GLM-562 tokens/s

Input price (USD per 1M tokens)lower is better

GPT-5 nano$0.05

Claude Haiku 5.5$0.10

GPT-6 Luna$0.10

GPT-4.1 nano$0.10

GPT-4o mini$0.15

GPT-5 mini$0.25

Claude 3 Haiku$0.25

GLM-5$1.00

Output price (USD per 1M tokens)lower is better

GPT-5 nano$0.40

GPT-4.1 nano$0.40

Claude Haiku 5.5$0.50

GPT-6 Luna$0.50

GPT-4o mini$0.60

Inkling 2 Small$1.00

MiniMax-M2$1.20

GLM-5$3.20

Where to compare Z.ai (Zhipu) GLM-5

Independent leaderboards and official resources for comparing this model yourself. Rankings change often; each site shows its own current data.

Leaderboards

General leaderboards

Official resources

Why Z.ai (Zhipu) GLM-5 matters for businesses

Artificial Analysis lists GLM-5 (Reasoning) at about 62 tokens per second and $1.00/$3.20 per 1M input/output tokens, and now points users to later GLM-5.x releases. A permissive license on a frontier-class model gives agencies a strong self-hosting option for coding agents.

Sources for this story

New Released (release notes) (opens in a new tab)
Z.ai Docs · docs.z.ai
Official
GLM-5 model card (opens in a new tab)
Z.ai on Hugging Face · huggingface.co
Official
GLM-5 model analysis (opens in a new tab)
Artificial Analysis · artificialanalysis.ai
Benchmark

Last verified: November 5, 2026. IA Compa summarizes public announcements for business readers. We are not affiliated with the companies mentioned. Always confirm details, pricing, and availability with the official source.

#llm
#zai
#glm
#open-weights
#mit-license

Z.ai releases GLM-4.5 and GLM-4.5-Air, open hybrid reasoning models for agents and coding

GLM-4.5 (355B total, 32B active) and GLM-4.5-Air (106B total, 12B active) offer thinking and instant modes, 128K context and native function calling, with open weights.

Black line drawing of a shaded right triangle on two arrow axes, with dashed guide boxes, three small coloured dots and the letters "Hy" in a serif font, on a light grey background.

Nov 4, 2026TencentModels

Tencent releases Hy4 with image understanding and computer use; open weights promised "in the coming days"

Hy4, the full release of Tencent's 770B-parameter model previewed in August, adds image understanding and computer use. It costs about $0.83/$2.50 per million tokens and scores 35 on the Artificial Analysis Intelligence Index. Tencent promises open weights soon; as of November 7 they are not out.

Image: Tencent

Oct 23, 2026Moonshot AIModels

Moonshot AI launches Kimi K3.1 Pro, then publishes its 2.8T-parameter open weights

Kimi K3.1 Pro is a post-trained upgrade of Kimi K3 with a 1M-token context, a faster Highspeed variant, lower API prices than K3, and weights released on Hugging Face on November 4, 2026.

Image: Moonshot AI

Z.ai releases GLM-5, an MIT-licensed open model for agentic engineering

Z.ai (Zhipu) GLM-5 at a glance

What changed with Z.ai (Zhipu) GLM-5