First Move  ·  AI Software  · 
Inside the labs — Wednesday morning, 17 June

Z AI's GLM-5.2 Tops Open-Weight Benchmarks with 1M-Token Context for Coding

Chinese lab Z AI's new open-weight model sets a benchmark, while OpenAI faces a multi-state investigation and Anthropic retires older models.

ShareWhatsAppXLinkedIn

The story

The AI software landscape saw notable shifts this week, with Chinese lab Z AI making waves with its new open-weight model. GLM-5.2, released on June 13 but widely reported on June 17, has claimed the top spot among open-weight models on the Artificial Analysis Intelligence Index. This model features a substantial 1 million token context window and is specifically optimized for coding and long-term agentic tasks, including dual thinking-effort levels to refine its coding capabilities. Its performance is described as on par with some proprietary models, marking a significant advancement in the open-source arena.

Meanwhile, regulatory scrutiny intensified for a major U.S. lab as a coalition of U.S. state attorneys general launched an investigation into OpenAI. The probe will examine issues related to user safety, advertising practices, model behavior, and data handling. This development underscores the growing focus on accountability as AI adoption accelerates. Additionally, Anthropic officially retired its Claude Sonnet 4 and Claude Opus 4 models on June 15, a planned deprecation requiring users to migrate to newer versions like Opus 4.8.

Who moved

Z AI

What Changed: Released GLM-5.2, an open-weight model optimized for coding and agentic tasks, which achieved the highest score among open-weight models on the Artificial Analysis Intelligence Index.

Consequence: This positions Z AI as a significant player in the open-source AI community and offers developers a powerful new tool for complex software engineering workflows.

Sarvam AI

What Changed: Raised $234 million in the first close of its Series B funding round on June 15, valuing the Indian sovereign AI company at $1.5 billion, with HCLTech as the strategic lead investor.

Consequence: This substantial funding will support Sarvam's continued research on its next frontier model for agentic, coding, and cybersecurity use-cases, and expand its access to compute at scale.

OpenAI

What Changed: Became the subject of a multi-state investigation by a coalition of U.S. state attorneys general, focusing on user safety, advertising practices, model behavior, and data handling.

Consequence: This development adds to growing regulatory scrutiny on frontier AI labs, highlighting increasing concerns over the ethical and operational implications of advanced AI systems.

Anthropic

What Changed: Officially retired its Claude Sonnet 4 and Claude Opus 4 models on June 15, as part of the company's scheduled deprecation plan.

Consequence: This requires production teams still using these older model versions to migrate to newer, more capable offerings like Claude Sonnet 4.6 or Claude Opus 4.8.

NewCore

What Changed: Secured $66 million in a Seed funding round, announced June 15, led by Cyberstarts with participation from Index Ventures and Evolution Equity Partners.

Consequence: The funding will be used to build identity management solutions for AI agents acting as enterprise employees, signaling strong early conviction in AI agent infrastructure.

New models

GLM-5.2

Lab: Z AI

What: An open-weight model with a 1 million token context window, featuring dual thinking-effort levels.

Use: Optimized for coding and long-term agentic tasks, it aims to enhance software engineering capabilities and complex multi-step projects.

Market signals

Sarvam AI secured $234 million in the first close of its Series B funding round on June 15, at a post-money valuation of $1.5 billion, with HCLTech as the strategic lead investor.

Implication: This significant investment underscores strong investor confidence in full-stack, sovereign AI development, particularly in emerging markets like India, and highlights the potential for AI companies to achieve unicorn status.

A coalition of U.S. state attorneys general launched an investigation into OpenAI, examining issues related to user safety, advertising practices, model behavior, and data handling.

Implication: This signals an intensification of regulatory scrutiny on major AI labs, indicating that policymakers are increasingly focused on the governance and responsible deployment of advanced AI systems.

NewCore raised $66 million in a Seed round, announced June 15, to build identity management for AI agents.

Implication: This suggests a growing market demand for specialized security and operational infrastructure to support the increasing deployment of AI agents within enterprise environments.

What we'll be watching

Reporting + analyst voices: grounded via Google Search at publish time.