GLM-4.6V-Flash
Z.ai's 9B efficient vision model — native tool calling with MIT-licensed weights.
About
GLM-4.6V-Flash is the efficiency variant of Z.ai's GLM-4.6V vision-language series, released December 2025. At 9B parameters it is the small, fast sibling aimed at edge computing and real-time applications, distributed — like the full 4.6V — under the MIT license with weights on Hugging Face and code on GitHub.
The series' headline feature is native function calling in a vision model: it can invoke tools like search, image cropping, or chart recognition directly against visual inputs, eliminating the lossy intermediate text-conversion step most VLMs need. It carries a 128K-token context, an AIMv2-Huge vision encoder with an MLP projector, arbitrary aspect ratios up to 200:1, and video-frame input with explicit timestamp tokens for temporal reasoning. Z.ai claims state-of-the-art results across 20+ benchmarks (provider-claimed). Access is via Hugging Face weights, Z.ai's OpenAI-compatible API, and a web demo.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
Provider-reported: SOTA across 20+ benchmarks.
Independent: no third-party reproduction found — covered by VentureBeat and press at launch.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Repository verified live on Hugging Face (HTTP 200); MIT license; launch covered by VentureBeat and multiple outlets with consistent specs. Docked for: (1) benchmark claims are provider-run with no independent reproduction; (2) most coverage describes the 4.6V series jointly, with less Flash-specific detail.