Search

Z.ai: GLM-4.7 dominates open-source and IPO approaches

Z.ai: GLM-4.7 dominates open-source and IPO approaches

Z.ai enters history

Z.ai (Zhipu AI), often nicknamed “China’s OpenAI”, is set to make AI history with two major announcements: the launch of GLM-4.7, the new global open-source number one, and a historic IPO scheduled for January 8, 2026.


IPO: first AI-native LLM company to go public

December 30, 2025 — Z.ai announces its IPO in Hong Kong, becoming the first AI-native company specializing in LLMs to be listed on global markets.

Z.ai is set for its IPO on Jan 8, 2026. This journey has been powered by our developers, researchers, and users from Day 1.

🇬🇧 Z.ai is set for its IPO on Jan 8, 2026. This journey has been powered by our developers, researchers, and users from Day 1.@Zai_org on X

Why it’s historic

AspectDetail
DateJanuary 8, 2026
Stock ExchangeHong Kong
FirstFirst AI-native LLM company listed
Position3rd LLM player in China (after Baidu, Ali)

GLM-4.7: the new king of open-source

December 22, 2025 — Z.ai unveils GLM-4.7, a model that redefines open-source standards by surpassing leading proprietary models.

GLM-4.7 is featured on Artificial Analysis Intelligence Index, positioned as a leading open-source model.

🇬🇧 GLM-4.7 is featured on Artificial Analysis Intelligence Index, positioned as a leading open-source model.@Zai_org on X

Impressive benchmarks

BenchmarkGLM-4.7vs Predecessor
AIME 202595.7%
HMMT Feb. 202597.1%
SWE-bench Verified73.8%+5.8%
SWE-bench Multi66.7%+12.9%
Terminal Bench 2.041%+16.5%
τ²-Bench87.4%
HLE (with Tools)42.8%+12.4%

Comparison with giants

GLM-4.7 measures itself directly against GPT-5.1-High, Claude Sonnet 4.5, Gemini 3.0 Pro and DeepSeek-V3.2, displaying competitive or superior performance on the majority of reasoning and coding tasks.


Outperforms GPT-5.1 on Vending-Bench2

December 27, 2025 — GLM-4.7 becomes the first “profitable” open-source model on the Vending-Bench2 benchmark, specializing in long-horizon agentic tasks.

GLM-4.7 outperforms GPT-5.1 and stands out as the leading open-source model on Vending-Bench2, demonstrating strong capabilities in long-horizon agentic tasks.

🇬🇧 GLM-4.7 outperforms GPT-5.1 and stands out as the leading open-source model on Vending-Bench2, demonstrating strong capabilities in long-horizon agentic tasks.@Zai_org on X


#2 on Design Arena

December 25, 2025 — GLM-4.7 reaches second place on Design Arena, just behind Gemini 3 Pro, confirming its capabilities in UI generation and web design.

Ranking #2 on Design Arena. Thanks to @Designarena for showcasing our progress in aesthetics.

🇬🇧 Ranking #2 on Design Arena. Thanks to @Designarena for showcasing our progress in aesthetics.@Zai_org on X


December 24, 2025 — GLM-4.7 dominates the Hugging Face trending ranking again.

Trending #1 on @huggingface again! We’ll keep open-sourcing next year.

🇬🇧 Trending #1 on @huggingface again! We’ll keep open-sourcing next year.@Zai_org on X


Three modes of thinking

GLM-4.7 introduces a flexible reasoning architecture with three modes:

ModeDescription
Interleaved ThinkingIntegrated reasoning before each response
Preserved ThinkingPreservation of reasoning between exchanges
Turn-level ThinkingControl of reasoning at each dialogue turn

Aggressive pricing

Z.ai positions GLM-4.7 as an economical alternative to proprietary models:

AspectGLM Coding Plan
Price1/7th of competitors
Quota3x standard usage
AccessAPI Z.ai, OpenRouter
Open-sourceHuggingFace, ModelScope

Day-0 Integrations

GLM-4.7 benefits from immediate support on numerous platforms:

  • Fireworks AI : Day-0 support
  • Roo Code : Native integration
  • Kilo Code : Full support
  • Blackbox CLI : Agent available
  • Factory AI : Droid integration

Timeline December 2025

DateEvent
22 DecGLM-4.7 Launch
23 DecFireworks Day-0 support
24 DecTrending #1 Hugging Face
25 Dec#2 on Design Arena
27 DecOutperforms GPT-5.1 on Vending-Bench2
28 Dec#1 open-source Artificial Analysis
30 DecIPO Announcement - Jan 8, 2026

Access


Sources