Z.ai / Zhipu · Coding and reasoning models
Using GLM-5.2 with Codex and Claude Code
Zhipu's 744B open-weights model scored 62.1 on SWE-bench Pro against GPT-5.5's 58.6, at roughly a sixth of the cost — the strongest case yet for running a coding agent on something other than a US frontier model.
Not wired up
Not yet. We don't route to GLM today, and we'd rather say that than sell a plan that quietly doesn't cover it. If enough people ask, it's the first one we'd add: the weights are MIT-licensed and the API speaks the OpenAI wire format, so the work is small.
The numbers
- Built by
- Z.ai / Zhipu
- Released
- Jun 16, 2026
- Context window
- 1,000,000 tokens
- Per 1M tokens (in / out)
- $1.40 / $4.40
- In HaryAI
- Not wired up
List price from the vendor's own API — cache-miss input and output. Third-party hosts are often cheaper, and cached reads change the real number.
What it takes to use it
Z.ai's endpoint is OpenAI-compatible, so anything with a configurable base URL will talk to it — Codex through ~/.codex/config.toml, Claude Code through its own provider settings. The friction isn't the connection, it's the details: tool-calling schemas and long-context behaviour differ enough from GPT that agent loops built against OpenAI need a pass before you trust them on real work. Zhipu also sells a coding plan from about $10 a month, which beats per-token pricing if you run it hard.
Want this one?
Leave an email and we'll tell you if we wire it up. That's the whole message — there's no newsletter attached to it.
Figures checked 2026-08-06 against vendor pricing pages and launch coverage. Models re-price often — if something here is stale, tell us.