๐Ÿ•ท๏ธ
SetupAI
๐ŸŒStalecollected in 2h

Zhipu Shares Surge Amid Anthropic Market Gap Bets

Zhipu Shares Surge Amid Anthropic Market Gap Bets
PostLinkedIn
๐ŸŒRead original on The Next Web (TNW)
#geopolitics#market-trends#llm-competitionzhipu-ai

๐Ÿ’กGeopolitical shifts are driving massive valuation gains for Chinese AI labs like Zhipu.

โšก 30-Second TL;DR

What Changed

Zhipu shares surged up to 48% on Monday

Why It Matters

This market movement highlights the geopolitical fragmentation of the AI industry, where regional leaders are gaining valuation based on restricted access to Western models.

What To Do Next

Evaluate Zhipu's GLM API performance compared to Western models if your application requires deployment in the Chinese market.

Who should care:Researchers & Academics

๐Ÿง  Deep Insight

Web-grounded analysis with 23 cited sources.

๐Ÿ”‘ Enhanced Key Takeaways

  • โ€ขThe US government imposed export controls on Anthropic's advanced AI models, "Claude Mythos 5" and "Fable 5," citing national security concerns over potential 'jailbreaking' to identify software vulnerabilities and suspicions of access by a China-linked group.
  • โ€ขZhipu (internationally branded as Z.ai) recently launched its GLM-5.2 model, featuring a 1-million-token context window, which is slated for open-source release and offered at a significantly lower price point (one-tenth) compared to Anthropic's premium tiers.
  • โ€ขZhipu AI successfully completed its Initial Public Offering (IPO) on the Hong Kong Stock Exchange in January 2026, marking it as the world's first pure-play large language model developer to go public, and its stock has experienced an approximate 820% surge since its listing.
  • โ€ขThe US export control directive on Anthropic's models also mandated the suspension of access for foreign nationals, including Anthropic's own employees, which led the company to globally disable access to these specific models.
  • โ€ขZhipu is recognized as one of China's leading 'AI tigers' and a significant player in the domestic LLM market, often outperforming established internet giants on specialized benchmarks and holding an estimated 18% share of domestic enterprise LLM adoption in early 2026.
๐Ÿ“Š Competitor Analysisโ–ธ Show
Feature/MetricZhipu AI (Z.ai) GLM-5.2 / GLM-5 / GLM-4.7 / GLM-4.6 / GLM-4.5Anthropic Claude Opus 4.5 / 4.6 / Mythos 5 / Fable 5DeepSeek-V3.2 / DeepSeek-R1Moonshot AI Kimi K2.5Alibaba Qwen3.5ByteDance Doubao
Latest ModelGLM-5.2 (June 2026)Mythos 5 / Fable 5 (June 2026 - restricted)DeepSeek-V3.2 / DeepSeek-R1 (June 2026)Kimi K2.5 (June 2026)Qwen3.5 (June 2026)Doubao (June 2026)
Context WindowGLM-5.2: 1M tokens; GLM-4.7: 200K tokens; GLM-4.6: 205K tokensNot specified for Opus 4.5/4.6; Mythos/Fable restrictedNot specifiedKnown for long-contextNot specifiedNot specified
Parameters (Total/Active)GLM-4.7: ~400B; GLM-4.5: 355B/32BNot specifiedDeepSeek-V3.2: 385B (MoE)Not specifiedNot specifiedNot specified
ArchitectureMoE (GLM-4.5); Depth over widthNot specifiedMoE (DeepSeek-V3.2)Not specifiedNot specifiedNot specified
Key StrengthsCoding, Agentic tasks (GLM-5, GLM-4.5); Mathematical reasoning (GLM-4.7); Multimodal (GLM-4.5V, GLM-4.6V)Safety, general reasoning (prior models)Cost-efficiency, performanceAgentic tasksMultilingual (Chinese, Japanese, Korean)Most popular chatbot in China (100M+ DAU)
BenchmarksGLM-5: SOTA in Coding/Agent, on par with Claude Opus 4.5; GLM-4.5: Ranks 3rd overall on 12 benchmarks, ahead of Claude 4 Opus, GPT-4.1, Gemini 2.5 Pro in coding; GLM-4.7: 73.8% SWE-bench Verified, 95.7% AIME 2025Claude Opus 4.5 (benchmark for GLM-5)DeepSeek-R1: comparable to US modelsKimi K2.5 leads on agentic tasksQwen3.5 strong in multilingualNot specified
Pricing (API)GLM Coding Plan: ~1/10th of Anthropic's premium tiers; GLM-4.7: $0.10/1M tokens (input/output); GLM-4.6: $0.400/1M input, $1.74/1M outputOpus 4.6: higher price than Z.ai's APIDeepSeek-V3.2: $0.28/1M input, $0.42/1M outputNot specifiedNot specifiedDoubao subscription tiers (68-500 yuan/month)
Open-Source StatusGLM-5.2 (upcoming); GLM-4.5 (MIT-licensed)Not specifiedDeepSeek-R1 (open-source)Not specifiedQwen (open-source)Not specified
Hardware SupportHuawei Ascend, Cambricon, Moore Threads GPUsNot specifiedNot specifiedNot specifiedNot specifiedNot specified

๐Ÿ› ๏ธ Technical Deep Dive

  • Zhipu's flagship product is the GLM (General Language Model) family of large language models, which are often released under the free and open-source MIT License.
  • The GLM training algorithm, introduced in May 2022, utilizes an "autoregressive blank infilling" strategy to regenerate randomly removed segments of input text.
  • Zhipu operates a vertically integrated AI stack built around its bilingual GLM architecture, designed to optimize training efficiency and reduce inference costs.
  • GLM-4.5 models employ a Mixture-of-Experts (MoE) architecture and a dual-mode system that switches between a "thinking" mode for complex reasoning and tool use, and a "non-thinking" mode for faster responses.
  • The architecture of GLM-4.5 prioritizes depth over width, featuring 96 attention heads per layer, and incorporates techniques like QK-Norm, Grouped Query Attention, Multi-Token Prediction, and the Muon optimizer for improved performance.
  • Training for GLM-4.5 was conducted on a 22-trillion-token corpus, with 7 trillion tokens specifically dedicated to code and reasoning, utilizing Zhipu AI's in-house asynchronous agentic reinforcement learning (RL) infrastructure.
  • GLM-4.6 models integrate FP8 and Int4 quantization on Cambricon chips and support native FP8 on Moore Threads GPUs, demonstrating a focus on domestic hardware compatibility.
  • GLM-4.7 is built upon approximately 400 billion parameters, supports a 200,000-token context window with a maximum output capacity of 128,000 tokens, and achieves an inference efficiency of 55 tokens per second.
  • The latest GLM-5.2 model boasts a massive 1-million-token context window.
  • Zhipu has developed multimodal capabilities, including the GLM-4.5V (a 106B parameter vision-language model) and GLM-4.6V, which focuses on visual understanding accuracy and comprehensive function call support.

๐Ÿ”ฎ Future ImplicationsAI analysis grounded in cited sources

Chinese AI firms like Zhipu will accelerate their global market expansion.
US restrictions on leading Western models create a vacuum and push users towards alternatives, especially given Zhipu's competitive pricing and open-source strategy.
The geopolitical divide in AI technology will deepen.
The US government's use of export controls on AI models signals a strategic shift to limit foreign access, prompting other nations to bolster domestic AI capabilities.
Zhipu's focus on domestic chip compatibility will drive innovation in China's AI hardware sector.
Zhipu's integration of models with Chinese chips like Huawei's Ascend and Cambricon incentivizes further development and adoption of indigenous AI hardware.

โณ Timeline

2019
Zhipu AI founded as a spin-off from Tsinghua University.
2022-05
Researchers published a paper introducing the GLM (General Language Model) training algorithm.
2023
Raised 2.5 billion yuan (approx. $350M USD) from major investors including Alibaba and Tencent.
2024-05
Secured $400M in a financing round, valuing the company at approximately $3 billion USD.
2026-01-08
Zhipu AI (Knowledge Atlas Technology) held its IPO on the Hong Kong Stock Exchange, becoming the world's first pure-play LLM developer to go public.
2026-06-15
Zhipu AI's shares surged following the release of its GLM-5.2 model, coinciding with US restrictions on Anthropic's advanced AI models.
๐ŸŒRead original article on The Next Web (TNW)
๐Ÿ“ฐ

Weekly AI Recap

Read this week's curated digest of top AI events โ†’

๐Ÿ‘‰Related Updates

Same topic

Explore #geopolitics

Same product

More on zhipu-ai

Same source

Latest from The Next Web (TNW)

๐Ÿ“Š

Prosus Profits Surge on E-commerce and Tencent Growth

Bloomberg Technologyโ€ขJun 19
Chinaโ€™s Race to Build Trillion-Parameter AI Models

Chinaโ€™s Race to Build Trillion-Parameter AI Models

SCMP Technologyโ€ขJun 19
Apple opens iOS to third-party app stores in Brazil

Apple opens iOS to third-party app stores in Brazil

The Next Web (TNW)โ€ขJun 18
General Intuition raising $300M to train AI on games

General Intuition raising $300M to train AI on games

The Next Web (TNW)โ€ขJun 18

AI-curated news aggregator. All content rights belong to original publishers.
Original source: The Next Web (TNW) โ†—