GLM 5.2: A Million-Token Window at the Same Price

What shipped

GLM 5.2's headline change is a context length extended to one million tokens, which required a small architecture change. The model had quietly been available in the subscription for a few days, but the blog post, the open weights, and the API release all landed together. Pricing did not move — it's still $1.4 / $4.4 per million tokens, and the weights are MIT-licensed.

GLM 5.1 figures shown for comparison; the 5.2 numbers are from the release. Run your own evals before trusting any vendor benchmark.

Why a 1M window at flat pricing matters

A million-token window means you can load an entire monorepo, a full contract set, or weeks of an agent's history in a single call — and because the price didn't increase, that headroom is effectively free. Open MIT weights mean you can also self-host if you need to. For an always-on agent, more context at the same rate is a direct upgrade to how much it can reason over at once.

Run it on MegaBrain

GLM 5.2 is already in the MegaBrain catalog — no waiting, no markup:

import openai

client = openai.OpenAI(
    base_url="https://getmegabrain.com/api/gateway/v1",
    api_key="mb-your-key-here",
)

response = client.chat.completions.create(
    model="z-ai/glm-5.2",
    messages=[{"role": "user", "content": "Read this whole repo and find the dead code"}],
)

See live pricing and every available model on the models page.