Z.ai (formerly Zhipu AI)
Updated May 2026
GLM
Beijing's university-spun open-weights flagship
Z.ai was spun out of Tsinghua University and ran as Zhipu AI before rebranding in July 2025. In January 2026 they IPO'd on the Hong Kong Stock Exchange. GLM-5.1 is the first open-source model to lead SWE-Bench Pro, and notably, it was trained entirely on Huawei Ascend 910B chips — zero NVIDIA hardware. Weights ship under MIT license, the most permissive license in wide use.
Variants
4
Best at
MIT license (most permissive)
Max context
128K
tokens
Pricing
GLM-5.1 API
$0.95 in / $2.50 out
per million tokens
Self-host
$0
per MIT-licensed weights
Variants
Sort the table by context window or cost to find the right variant. Click any version below for a battle card with ranks, pricing notes, and official links.
| Modalities | ||||
|---|---|---|---|---|
GLM-5.1 glm-5-1 | 128K | varies / varies | 2026 | textcode |
GLM-5 glm-5 | 128K | varies / varies | 2026 | textcode |
GLM-4.6 glm-4-6 | 128K | varies / varies | 2025 | textvision |
ChatGLM-6B chatglm-6b | 8K | varies / varies | 2023 | text |
Battle card
Context rank
#1
within GLM
Capability rank
#1
modalities + reasoning
Weights
Open
self-hostable if licensed
Best fights to pick
- autonomous multi-hour coding
- open-source agent engines
- research benchmarking
Rankings are Tendril directory ranks, computed from the model data shown here. Public benchmark leaderboards change often, so official docs and current benchmark pages should be checked before buying or deploying.
Learn
Lessons about this model
Structured lessons that cover GLM directly or put it in context alongside its rivals.
Check yourself
Quizzes
Short, mixed-difficulty quiz sets on GLM and its model family.
Other Players: MiniMax, Cohere, Amazon, 01.AI, and More
5 questions
The broader landscape of model providers.
Start quiz →Hands-on
Try these prompts
Ready-made prompts that show GLM at its best. Use them in your own AI workspace, then compare the output with what you learned in Tendril.
GLM on a bilingual coding task
BuildersZhipu's GLM series handles mixed-language prompts well — ideal for cross-border engineering teams.
Convert this Chinese-commented JavaScript function into TypeScript with English comments, keeping the original logic identical.
function 计算总价(items) { /* 计算含税总价 */ ... }GLM reasoning on a classic puzzle
BuildersTest GLM's chain-of-thought on a math problem against DeepSeek R1.
Three friends split a restaurant bill three ways and each paid $10. The waiter realizes the bill was $25 and returns $5. Each friend keeps $1 and tips the waiter $2. So each friend paid $9, times 3 is $27, plus the $2 tip is $29. Where's the missing dollar? Explain.
Summarize a Chinese news article in English
BuildersTest cross-lingual summarization.
Summarize this Chinese news article in 3 English bullets with no loss of key numbers or names. [Paste article]
Print & keep
Printable reference
One-page summaries and flowcharts — great for desks, classrooms, or study sessions.
Go deeper
Official resources
Straight from the lab — docs, API references, and the chat surfaces you can try today.
Strengths
- MIT license (most permissive)
- trained on non-NVIDIA hardware — proves export-control workaround
- competitive SWE-Bench performance
Limits
- PRC censorship
- less international brand recognition than DeepSeek
- Chinese-first documentation