You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Browse filesBrowse the repository at this point in the historyBrowse files
鲁工
authored and
鲁工
committed
feat: add GLM-5.3-Flash and Qwen3.8 Flash, v1.16.0
Two new multimodal flash-tier models, added as variants on the existing
providers so no new keys or endpoints are introduced; defaults unchanged.
GLM-5.3-Flash (glm-plan-5.3-flash, glm-global-5.3-flash): first native
multimodal GLM-5 model, 1M context / 128K max output, thinking cannot be
disabled. Both vendor Claude Code guides (docs.bigmodel.cn coding-plan/
tool/claude, docs.z.ai devpack/tool/claude) configure glm-5.3-flash on
the same Anthropic endpoints already used by glm-plan / glm-global.
Specs: docs.bigmodel.cn/cn/guide/models/vlm/glm-5.3-flash and
docs.z.ai/guides/vlm/glm-5.3-flash. Unverified upstream: no GLM keys.
Qwen3.8 Flash (qwen3.8-flash, qwen-plan-3.8-flash): 1M context /
131,072 max output, image+video input; listed in Bailian's Anthropic-
compatible Messages API model set (help.aliyun.com/zh/model-studio/
anthropic-api-messages). Pay-go verified: `ccmr doctor qwen3.8-flash`
-> [OK] 1.69s. Token Plan variant kept on the strength of the
platform.qianwenai.com latest-model doc ("Token Plan supports
Qwen3.8-Flash"); the overview tier tables do not list it but are stale
(still show retired glm-5.1). Not verified live: the test subscription
returns 403 AccessDenied.Unpurchased for the shipped 3.8-max as well.
qwen3.8-flash-next is deliberately NOT added: it is the open-weights
Qwen4-architecture preview (HF Qwen/Qwen3.8-Flash-Next); Bailian 404s on
it and no first-party catalog or OpenRouter serves that id; the HF card
names qwen3.8-flash as the production API. A test pins its absence. The
international Qwen line (dashscope-intl / QwenCloud) exists but is out
of scope by user decision.
0 commit comments