SHELFCRITTER AI Digest

COMPANY

Zhipu AI

companytopic-notechina

Overview

Zhipu AI (also known as Z.ai on its international product surface) is a Chinese frontier lab and the creator of the GLM 5.2 open-weights model family. Through mid-2026 Zhipu has moved from a weights-only distribution posture to a full-stack coding-agent product play, launching ZCode on July 7 as an explicit competitor to Claude Code and OpenAI Codex — 1M-token context, aggressive trial economics (5M free tokens/day for five days plus paid plans starting $18/month), and API pricing roughly 1/6th of GPT-5.5.

Timeline

  • 2026-07-08-AI-Digest — Zhipu AI shipped ZCode, a GLM 5.2-powered coding agent positioning explicitly against Claude Code and OpenAI Codex — 1M-token context, five-day new-user trial of 5M free tokens per day (3M GLM 5.2 + 2M GLM-5-turbo), paid plans starting $18/month, API pricing at $1.40 / $4.40 per M in/out (~1/6th of GPT-5.5). The Decoder cites a Snowflake CEO write-up of a 103-task dbt-bench comparison in which GLM 5.2 and Claude Opus 4.7 land 66% vs 67% at Pass@3 — but with a wider first-attempt gap (47.6% vs 53.7%) and roughly 2× the token usage on the GLM 5.2 side. Narrow read: on one SQL-coding benchmark at three attempts GLM 5.2 is near-parity, but the Pass@1 gap and 2× token cost tell a different story about single-shot reliability and inference economics. Structural read the digest carries: the pricing is the news, not the benchmark — the most aggressive coding-agent economics any Chinese lab has taken to market against Claude Code — pairs with today’s Tencent Hy3 open-weights release as two independent pressure points on the coding-agent cost stack in one week.

Key Developments

  1. ZCode Coding Agent Launch (July 8, 2026): First full-stack coding agent from a Chinese lab positioned explicitly against Claude Code and Codex — 1M-token context, five-day trial at 5M tokens/day, $18/mo paid plans, ~1/6th GPT-5.5 API pricing. The pricing is the load-bearing news; the GLM 5.2 / Claude Opus 4.7 66%/67% Pass@3 dbt-bench near-tie is a single-benchmark result at three attempts with a 2× token cost, not a general parity claim.

  2. GLM 5.2 as Foundation: ZCode runs on Zhipu’s GLM 5.2 plus a GLM-5-turbo fallback for trial tier — the coding-agent product surface consolidates Zhipu’s open-weights model line under a first-party product rather than leaving distribution to third-party wrappers.

See also: GLM 5.2, ZCode, Z.ai, Claude Code, MOC - Open Source Models, MOC - Major Companies.