Breaking

Cloudflare Adds Z.ai's GLM-5.2 to Workers AI, Runnable on Free Tier

June 28, 2026 at 05:06 EDT

Foundation Models
AI Agents
Software Dev & Coding

On June 16, 2026, Cloudflare added GLM-5.2, an agentic coding model from Z.ai (Zhipu AI), to its serverless inference platform Workers AI. It connects to existing developer tools via an OpenAI-compatible endpoint, and the ability to run a coding agent at zero cost within a daily free quota is drawing attention.

June 16, 2026 · Cloudflare Workers AI

GLM-5.2 Lands on Workers AI — Run Coding Agents on a Free Tier

Cloudflare now officially hosts Z.ai's flagship agentic coding model on its serverless GPU platform. An OpenAI-compatible endpoint plugs straight into Cursor, Aider, OpenCode, and Claude Code — with a daily free quota that has developers paying attention.

744B-A40B

Mixture-of-Experts scale, MIT open weights

262K

Token context window (1M planned)

10K

Free neurons per day (UTC reset)

What a token costs in neurons

Output burns roughly 3× the compute of input (per 1M tokens). Same unit, drawn to scale.

127k

neurons /
1M input

400k

neurons /
1M output

The daily 10k free neurons ≈ 70k–80k input tokens and ~25k output tokens — fine for light use, quick to exhaust on big tasks.

Input

$1.40 /1M

Output

$4.40 /1M

Cached input

$0.26 /1M

Overage

$0.011 /1k neurons

Welcomed for

Try coding agents for free
Drops into Cursor, Aider, OpenCode, Claude Code
Repo-level coding, UI generation, tool integration

Caution

10k neurons/day suits light usage only
Large tasks hit the limit fast — not unlimited
Overage is paid; use caching for efficiency

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.