Zhipu AI releases GLM-5.2, tops open-weight rankings

Zhipu AI released GLM-5.2, a mixture-of-experts model with 744 billion total parameters and around 40 billion active per token, a 1-million-token context window, and no regional restrictions on the MIT licence covering its weights. On coding benchmarks, Zhipu reported 81.0 on Terminal-Bench 2.1 and 62.1 on SWE-bench Pro, and said the model beat OpenAI’s GPT-5.5 on several long-horizon coding benchmarks at roughly a sixth of the per-token cost, while landing within about a percentage point of Anthropic’s Claude Opus 4.8 on FrontierSWE. On Artificial Analysis’s Intelligence Index, GLM-5.2 scored 51, the highest recorded for an open-weight model at the time. Zhipu priced API access at $1.40 per million input tokens and $4.40 per million output tokens.

The release landed a day after the US Commerce Department ordered Anthropic to take its Fable 5 and Mythos 5 models offline for foreign users worldwide over an export-control dispute, and commentary around GLM-5.2 explicitly framed openly licensed Chinese models as a way to route around exactly that kind of access restriction: weights published without regional limits cannot be withdrawn from a jurisdiction the way a hosted API can. That framing reinforced a wider 2026 pattern in which the strongest open-weight coding models were coming disproportionately from Chinese labs, competing on price and availability against closed, more heavily export-controlled US frontier models.

GLM-5.2 went on to become the top-ranked open-weight model on independent leaderboards within the following weeks, though at a price high for its class relative to other Chinese open releases.