Z.ai: GLM 5.1

Four consecutive main-line GLM releases published the same context window. z-ai/glm-4.6 lists 204800tokensopenrouter-models, last checked 2026-09-14, z-ai/glm-4.7 lists 204800tokensopenrouter-models, last checked 2026-09-14, z-ai/glm-5 lists 204800tokensopenrouter-models, last checked 2026-09-14, and this row lists 204800tokensopenrouter-models, last checked 2026-09-14 — the same figure on all four in the 31 August 2026 snapshot, spanning the six months from 30 September 2025 to 7 April 2026. Whatever Z.ai spent those six months on, it was not the window.

Some of it went onto the price. z-ai/glm-5, listed 11 February 2026, heads at $0.60per million tokensopenrouter-models, last checked 2026-09-14 for input; this row, eight weeks later, heads at $0.97per million tokensopenrouter-models, last checked 2026-09-14 in the same snapshot, for an identical window and weights that stay open: MITsource, accessed 2026-08-28 over 754B totalsource, accessed 2026-08-28, per the model card for zai-org/GLM-5.1. Each of those is the top listed provider's rate for its row rather than necessarily Z.ai's own, and the two rows are headed by different providers, so the distance between them measures the hosts as much as the models.

Then the plateau broke. Ten weeks after this row, z-ai/glm-5.2 arrived listing 1048576tokensopenrouter-models, last checked 2026-09-14 of context — the largest single step anywhere in the GLM line — while its input listing stayed beside this row's instead of climbing with the window: $0.68per million tokensopenrouter-models, last checked 2026-09-14 there against $0.97per million tokensopenrouter-models, last checked 2026-09-14 here. Those two figures come from different top providers, so how they line up is a fact about hosts rather than about the models; what the pair does show is that a fivefold window did not arrive with a fivefold price. This row is the last of the main line on the old envelope.