A new flagship AI coder just landed. GLM-5.2 from Z.ai — 1M-token context, frontier-level coding — and you get it on a flat monthly Coding Plan, not a per-token meter. I plugged it straight into my Agent OS. Then I told it to build things. It did. Press play below.
Every tile below is a real, single-file build GLM-5.2 wrote, running in my Agent OS. Click any one and it plays. No editor. No assets to wire up. Just a prompt in, a working thing out.
On 13 June 2026, Z.ai released GLM-5.2 to everyone on their Coding Plan — every tier, Lite to Team.
It's their new flagship model. Strong coding. A usable 1-million-token context window. And it holds up on long, multi-step jobs — the kind where most models lose the plot halfway through.
It has two thinking levels: High and Max. For real coding, you run Max — it thinks harder before it writes, and the builds come out cleaner.
One honest note: it's text-only right now. No image input yet. For coding and building, that's fine. The standalone API, the chatbot, and the open-source MIT release are all coming right after launch.
"GLM-5.2 is now available to all GLM Coding Plan users… As our new flagship model, GLM-5.2 delivers powerful coding capabilities, usable 1M-context support, and continued strengths in long-horizon tasks."
— Z.ai (@Zai_org), 13 June 2026
Because it's frontier-level coding on a flat plan instead of a per-token bill that drains while you build.
And it runs inside the agents you already use — Claude Code, Cline, Hermes, and now a GLM section in Agent OS.
Everything here comes from Z.ai's own announcement and docs — nothing second-hand. Here's where to read it and switch your own agent to GLM-5.2:
The OpenAI-compatible base URL for any coding agent is https://api.z.ai/api/coding/paas/v4, model glm-5.2. In Claude Code, add the [1m] suffix for the full 1M window.
Before
I was paying per token for top-tier coding.
Every big build I asked for cost real money.
I'd be mid-build and the balance would run dry, or I'd get rate-limited.
So I'd hold back — give the AI less context, ask for smaller jobs, to save spend.
I was rationing the exact thing that makes it good.
Then I put GLM-5.2 on a flat Coding Plan into Agent OS.
After
Now I hand it the whole job and its full 1M context.
I run it on Max effort without watching a meter.
It builds real apps and games in one shot — they show up playable in my dashboard.
One predictable price. No balance to babysit.
You can have this too. Same plan. Same path.
Here's what's happening for the members already running this stack — agency owners, ecom founders, course creators, solo operators. Different businesses. Same engine room.
You've seen the builds. You've seen the proof.
The next few minutes show exactly how I wired GLM-5.2 in.
So here's the deal.
Promise yourself one thing right now. You finish this guide, and you switch one of your coding agents to GLM-5.2 before you sleep tonight. Just one. Because the moment you stop rationing a frontier coder by the token, the size of what you build changes.
The people sitting still are still watching a meter. The people who switch today are the ones shipping whole apps in a sitting.
Be one of those people.
Commit to the transition. Take action today. This changes how much you can build.
A frontier-class coder you stop rationing. Five plain pieces — that's the whole idea: a genius brain you can finally let off the leash.
You stop paying per token. One predictable monthly price means you never hold back a build to save spend again.
You hand it the whole job. A 1-million-token window holds your full codebase, so it stops forgetting what it's building halfway through.
You flip on deep thinking for the hard stuff. It reasons longer before it writes — so the first version actually works.
You run it everywhere. The same plan powers Claude Code, Cline, Hermes — and a dedicated GLM section in your Agent OS.
You get real things, not chat. Apps, games and tools land as finished, playable files you can open and ship.
It's their flagship, not their budget model. The builds on this page are the proof — real, working, one-shot.
Flat-plan pricing is about how you pay, not how good it is.
I added GLM-5.2 to my Agent OS as its own thing — a "GLM 5.2" item in the left nav.
Open it and there's a chat that talks straight to GLM-5.2 on the Coding Plan, and a workspace tab that shows everything it builds.
Ask it to build a game in the chat. A minute later it's a playable tile in the workspace. Click it. It runs.
You pick GLM-5.2 in a dropdown and type what you want. That's it.
The plumbing is already done inside Agent OS. Members who'd never opened a terminal are running agents like this.
Most models forget. You paste half your code, they lose the other half, and the answer breaks something three files over.
GLM-5.2 takes a million tokens of context. That's your whole codebase in one go.
So when you ask for a change, it already knows every file it touches. Fewer "oops, that broke the login" moments.
"The difference isn't a smarter sentence — it's that it remembers the whole job."
— the point of a 1M windowIf you want GLM-5.2 running in one dashboard with all your other agents — the way you saw above — get the Agent Operating System inside the AI Profit Boardroom.
It connects Claude, Hermes, and your agents into one place. They share one memory. They know your business. So GLM-5.2 builds in your style, with your context, from the first prompt.
158 pages of members who already broke through these exact beliefs. Real businesses, real wins, documented.
Read the 158-page testimonials doc →Get on the Z.ai GLM Coding Plan (any tier works).
Grab your Z.ai API key from the dashboard.
Point your coding agent at https://api.z.ai/api/coding/paas/v4, model glm-5.2.
In Claude Code, add the [1m] suffix to unlock the full 1M context.
Set effort to Max for real coding jobs.
Or drop it into Agent OS — pick GLM 5.2 in the nav and just chat.
Ask for the whole build, not a slice. Feed it real context.
Open what it makes in the workspace. Ship it.
Frontier coding on a flat plan.
1M tokens — feed it the whole project.
Max effort thinks before it writes.
Same model in every agent + Agent OS.
Real playable apps from one prompt.
Building while everyone else waits.
A frontier coder. Off the leash.
If you want GLM-5.2 to actually save you time every day — not be one more tab — grab the Agent Operating System inside the AI Profit Boardroom.
It turns Claude, Hermes and your agents into one system with shared memory and one dashboard. GLM-5.2 slots right in, already knowing your business.
