the first natively multimodal model in Zhipu AI's GLM-5 series, an open-weights 320B-total, 18B-active sibling of GLM-5.3. it costs about a ninth of GLM-5.3's per-token rate and, unlike GLM-5.3, accepts image and video input.
pick it if
you want GLM-5.3-family text at roughly a ninth of the price
you need image or video input in the GLM line
you want MIT-licensed weights you can self-host
skip it if
you need replies without thinking; Z.ai says thinking cannot be disabled
Qwen3.8 Flash was Alibaba's August 2026 hosted flash-tier model, the production version of the open-weight Qwen3.8-Flash-Next (125B total, 6B active). it was retired on 2026-10-06 in favour of Qwen3.8 Omni Flash, which bills the same and accepts more input kinds.
pick it if
ran a 1M-token context at flash-tier prices
wanted cache hits billed at about a tenth of the input rate
needed image and video input in a low-cost tier
skip it if
you need it today: Qwen3.8 Omni Flash replaced it at the same price
faq
which is cheaper, GLM-5.3 Flash or Qwen3.8 Flash?
Qwen3.8 Flash lists the lower output price; input prices match. on a coding-agent session (2M tokens in, 60,000 out) GLM-5.3 Flash costs $0.33 and Qwen3.8 Flash $0.33.
try either model
switch between models inside the same octomind session.