Z.ai

GLM-5.3-Flash

Made by Z.ai (Zhipu AI). Released 2026-08-20. Accessibility not recorded. Cheapest standard rate we hold: $0.50 per million output tokens, no context split.

What it costs

Tier	Kind	In	Out	Cached in	From
standard	text	$0.15	$0.50	$0.03	seller

per million tokens, US dollars. * marks a rate from a single registry or two that disagree — the best available number rather than a confirmed one. Every other row is confirmed against a seller or a licensed index.

What it scores, by house

Benchmark	House	Score	Evaluated	Scaffold
chess puzzles	Epoch AI	0.2±0.04	2026-10-13	max
frontiermath tier 4 v2	Epoch AI	0.2±0.06	2026-08-27	max
frontiermath tiers 1 3 v2	Epoch AI	0.6±0.03	2026-08-27	max
gpqa diamond	Epoch AI	0.9±0.02	2026-08-26	max
mystery game puzzles	Epoch AI	0.1±0.03	2026-08-28	max
otis mock aime 2024 2025near its ceiling	Epoch AI	0.9±0.02	2026-08-26	max
ale bench	collected by Epoch AI	304	—	high
apex agents	collected by Epoch AI	0.7	—	unknown
arc aginear its ceiling	collected by Epoch AI	0.7	—	high
arc aginear its ceiling	collected by Epoch AI	0.9	—	max
arc aginear its ceiling	collected by Epoch AI	0.5	—	low
arc agi 2	collected by Epoch AI	0.3	—	low
arc agi 2	collected by Epoch AI	0.7	—	max
arc agi 2	collected by Epoch AI	0.5	—	high
btf3	collected by Epoch AI	0.1	—	FutureSearch agent
critpt	collected by Epoch AI	0.2	—	unknown
cursorbench	collected by Epoch AI	0.3	—	low
cursorbench	collected by Epoch AI	0.4	—	max
cursorbench	collected by Epoch AI	0.3	—	high
deepswe	collected by Epoch AI	0.6	—	mini-swe-agent
frontiercode	collected by Epoch AI	0.3	—	devin-cli
frontierswe	collected by Epoch AI	0.2	—	proximus
gdp pdf	collected by Epoch AI	0.1	—	max
posttrainbench	collected by Epoch AI	0.4	—	Claude Code
proofbench	collected by Epoch AI	0.2	—	max
scicode	collected by Epoch AI	0.5	—	unknown
surface evolver bench	collected by Epoch AI	0.5	—	max
webdev arena	collected by Epoch AI	1607	—	unknown

These scores are not comparable between houses and are not added up. A score is dated by when it was evaluated, not when the model was released, and a score without its scaffold is not a measurement — which is why both are printed.

Every rate we hold · Why there is no single best model · Everything from Z.ai