Introducing GLM-5V-Turbo: Vision Coding Model - Native Multimodal Coding: Natively understands multimodal inputs including images, videos, design drafts, and document layouts. - Balanced Visual and Programming Capabilities: Achieves leading performance across core benchmarks for multimodal coding, tool use, and GUI Agents. - Deep Adaptation for Claude Code and Claw... Scenarios: Works in deep synergy with Agents like Claude Code and OpenClaw. Try it now: API: Coding Plan trial applications:show more

The model can understand design drafts, screenshots, and web interfaces to generate complete, runnable code, truly achieving the goal of "seeing the screen and writing the code."

GLM-5V-Turbo leads in benchmarks for design draft reconstruction, visual code generation, multimodal retrieval and QA, and visual exploration. It also performs exceptionally well on AndroidWorld and WebVoyager, which measure control capabilities in real GUI environments.

Regarding pure-text coding, GLM-5V-Turbo maintains stable performance across three core benchmarks of CC-Bench-V2 (Backend, Frontend, and Repo Exploration), proving that the introduction of visual capabilities does not degrade text-based reasoning.

The leading performance of GLM-5V-Turbo stems from systematic upgrades across four levels: Native Multimodal Fusion: Deep fusion of text and vision begins at pre-training, with multimodal collaborative optimization during post-training. We developed the next-generation CogViT visual encoder, reaching SOTA in general object recognition, fine-grained understanding, and geometric/spatial perception. We also designed an inference-friendly MTP structure to ensure high efficiency. 30+ Task Collaborative RL: The RL stage optimizes over 30 task types simultaneously, covering STEM, grounding, video, and GUI Agents. This improves perception and reasoning while mitigating the instability often found in single-domain training. Agentic Data and Task Construction: To solve the challenge of scarce Agent data, we built a multi-level system ranging from element perception to sequence-level action prediction. We use synthetic environments to generate verifiable training data and inject "Agentic Meta-capabilities" during pre-training (e.g., adding GUI Agent PRM data to reduce hallucinations). Multimodal Toolchain Extension: Beyond text tools, the model supports multimodal search, drawing, and web reading. This expands the perception-action loop into visual interaction. Synergies with Claude Code and OpenClaw are enhanced to support full-loop task execution.

Will it be open sourced?

Weights wen?

Guys, I agree it’s nice but where is the finesse on the design ? Does this page on an iPhone 13Pro max look good to you? :(

sus

My Claude code sub about to end… and I don’t think I’ll be resubscribing to ask my 2 questions/day. Not after THIS exists

GLM 5.1 in my own agent harness runs circles around Opus 4.6 in Antigravity. I literally also built my own coding cli the day Anthropic reduced the quota in Antigravity to hilarious 4 complex coding tasks… for Ultra users that is. The Zai Max Plan is currently by far the best offer for everything coding or agent related. The new models are incredible!

I tested some of the multimodel functions in hermes agent! I was really impressed at how well it read from screenshots, handwriting, and messy charts. Fast too! Check out the results in this quick video:

Can’t wait to try it 💥

I learned the hard way that generating code from a mockup is never the actual bottleneck.

seems like knowledge cut off 2019🤣🤣🤣

Its sad to see Zai go so far down hill so fast. Its like after you went public you decided to join the Dr Evil AI lab crowd like Anthropic and OAI. First you served a broken model to your users for WEEKS and then you went back on your oss mission and aren't releasing that model

@grok GLM-5.1 vs GLM-5V-turbo for real coding, who is better and faster?

really cool!

Except that it's not available on pro accounts 😞.

It is now available! Give it a try❤️

this is what I see inside OpenCode.

Please open source it !

The multimodal coding approach is huge - being able to parse UI mockups and generate code directly from visual inputs cuts development time dramatically. This is where vision models really shine.

Source? Weights? 🙏🏽

so does this mean we dont need MCP for image anymore?

Hopefully! This model is good at SVG

This will rock with Hermes @Teknium

i hope its not april fools joke >.<

Can't wait to use this as my project manager in CAR

finally someone tackling multimodal coding properly. the visual understanding piece has been the missing link for agentic dev

How much better is 5.1 for non-visual tasks?

fuck im hard

Wow

@grok Should I replace GLM-5.1 with GLM-5V-Turbo?

@SookieWinehouse

Fix the GLM-5.1 inference server, first guys!

@n1d_all open weight wen

Buying 1 year subscription with new year discount was a beast deal.

"Deep Adaptation for Claude Code and Claw Scenarios: Works in deep synergy with Agents like Claude Code and OpenClaw." What are the precise OpenClaw examples? All I want is for Chinese LLMs to be on par with GPT 5.2 and Opus 4.5 as personal assistants, yet they still come short

Would you use Chinese models, now that they have all switched to closed source?