Mistral serves Z.ai GLM 5.3 in public preview with a 1M context at $1.40 per million input tokens

Sources: primary, discussion
Summary: The model card, dated 2026-09-15, states Mistral hosts the third-party open model without modifications, with a 1M context window and 128k maximum output. Published prices are $1.40 per million input tokens, $0.14 cached, and $4.40 per million output tokens, with function calling, structured outputs and batching listed as supported. The page gives no evaluation results.
Why it matters: Teams that cannot route traffic to Z.ai directly now have a European-hosted path to the same weights with published terms.