Chinese AI lab Zhipu has shipped GLM 5.2, a coding-first model now live across every tier of its Z.ai Coding Plan. The headline feature is a 1-million-token context window, roughly five times the window of GLM 5.1, with a standalone API and MIT-licensed open weights scheduled to follow the next week.

What This Enables

If you already pay for the GLM Coding Plan, you can select the new model today by pointing your client at the glm-5.2[1m] model ID. The 1M-token window means you can load an entire mid-sized repository, a long design spec, and your test files into a single prompt instead of chunking them across calls. GLM 5.2 also adds a dual thinking-effort control with High and Max settings, so you can trade latency for deeper reasoning on harder refactors. Output is capped at 131,072 tokens per response, enough to generate a full module or a large multi-file diff in one pass.

Why It Matters for Creators

For builders wiring AI into creative pipelines, context length is the constraint that decides whether a model can hold a whole project in view at once. A usable 1M window puts GLM 5.2 in the same bracket as the largest commercial context offerings, but on an open-weights track. When the MIT-licensed weights reach Hugging Face, teams will be able to self-host the model with no per-token API cost, which matters for anyone running high-volume agentic coding or generation jobs. That continues the same open-weights pressure we saw when Kimi K2.7-Code beat frontier models on tool use.

Key Details

Availability: Live now on all GLM Coding Plan tiers (Lite, Pro, Max, Team), which start at roughly $18 per month.

Context window: 1,000,000 tokens via the glm-5.2[1m] model ID, with up to 131,072 output tokens per response.

Reasoning: Dual thinking-effort modes, High and Max.

Coming next week: A standalone API, Z.ai chatbot access, and MIT open weights. Zhipu did not publish benchmark numbers at launch, so its performance claims are unverified for now.

What to Do Next

Coding Plan subscribers should switch one real task to glm-5.2[1m] and feed it more context than usual, a full repo or a long spec, to see whether the larger window changes the quality of its edits. If you are weighing a self-hosted setup, wait for next week's weights release and the first independent benchmarks before committing, since Zhipu has not shared performance data yet.