Cindy is live today — and open source.
Built to work out of the box, while staying yours to inspect, extend, and shape.
The launch model is also available to try for a limited time.
GitHub: https://t.co/33QMTr15QM
Yes, https://t.co/9c4OarfRe9 has been updated for the first time. 🐳DeepSeek supports multimodality and can understand pictures.
For models that do not support multimodality, Cindy supports opening the visual bridge.
Enjoy it🚀
DeepSeek-V4-Flash-Vision-Exp is now live on the DeepSeek API Platform! 🚀
🔹 This experimental multimodal model matches DeepSeek-V4-Flash on text capabilities—including agents, reasoning, and world knowledge.
🔹 On multimodal agent benchmarks, V4-Flash-Vision-Exp makes a major leap over V4-Flash, bringing multimodal agent performance close to Opus-4.8.
Try it with model='deepseek-v4-flash-vision-exp'. DeepSeek Harness 0.1.1 was released today with out-of-the-box support for the new model.
1/n
🪄 iOS Simulator is now part of Cindy Desktop!
You can view the screen, tap, type, and launch apps right inside Cindy — and let your Agent interact with the iOS environment along the way.
From writing code to seeing it come to life on iOS, without jumping between tools. 📱✨
Here’s a quick look at it in action 👇
#Cindy #iOS #AIAgent
The model race is officially on fire. 🔥
New contenders keep entering the arena, benchmarks change overnight.
This round:
GLM-5.3 goes deeper on complex coding and long-running tasks.
Gemini 3.7 Flash pushes harder on speed, cost, and high-volume work.
We ran a quick comparison in Cindy—but benchmarks are only a map, not the answer. The best model is the one that performs on your code, documents, and workflows.
Both models are now available in https://t.co/B3h6KZfNun.
Bring a real task. Try them side by side.
GLM-5.3 is now available on Cindy.
Built for stronger coding, terminal workflows💻, and long-horizon agent tasks🤖 —with more efficient token usage.
No additional setup required. Select GLM-5.3 from the model picker and put it to work on a real project.
The best benchmark is your own workflow. 👉https://t.co/K4wyqqo0iF
#CindyAI #GLM53 #CodingAgent #AIEngineering
Introducing GLM-5.3: Built to Code. Ready for Cyber Defense.
- Top-tier coding and agentic capabilities, achieved through post-training on the 743B base model
- A major leap in cybersecurity, setting a new standard among open models
Tech Blog: https://t.co/ekQkO83jCv
Cindy is live today — and open source.
Built to work out of the box, while staying yours to inspect, extend, and shape.
The launch model is also available to try for a limited time.
GitHub: https://t.co/33QMTr15QM
Grok 4.6 is now available in Cindy.
Built for sharper professional reasoning, more reliable long-running tasks, and stronger coding.
In third-party benchmarks vs. Grok 4.5:
• +10.4 pts on agent tasks 🤖
• +11.9 pts on DeepSWE ♻️
Select Grok 4.6 in Cindy and put it to work.
⬇️
https://t.co/K4wyqqnst7
Cindy at #Google IO today 👀
We care a lot about making AI feel native across devices.
That’s why Cindy has supported Android since day one — not as an afterthought, but as part of the core experience.
Desktop when you’re working. Mobile when you’re moving.
https://t.co/K4wyqqnst7
DeepSeek-V4-Pro-0813 is now available in Cindy
Cindy has moved from DeepSeek-V4-Pro-Preview to the latest DeepSeek-V4-Pro-0813 release.
The upgrade goes beyond higher scores. The biggest gains show up in demanding, real-world agent work:
🧠Long-horizon repository work: DeepSWE rises from 12.8 to 62.7.
🤖Difficult multi-step agent tasks: DSBench-Hard rises from 31.1 to 67.2.
🔐Vulnerability analysis and validation: CyberGym rises from 52.7 to 83.3.
💻Terminal and tool execution: Terminal Bench rises from 72.1 to 87.9.
https://t.co/K4wyqqo0iF
#Cindy #DeepSeek #AIAgent #CodingAgent
Cindy v0.1.35 / v0.1.36 are out 🚀
🚨 Workspace-capture concurrency fixed — same-directory tasks no longer stall behind a full turn
🧹 Empty +0 -0 review cards are gone
🧠 Live web search and a verified resource/process panel
🪟 Recursive task split views with up to eight panes, plus conversation-to-image sharing
🤝 Exportable Orca teams and cross-Agent/model task creation
🧩 Plugin compatibility, preferred models, Telegram history, and per-turn permission checks
🔗 Tighter model, Hook, Pi, IM, screenshot, and cross-device stability
20 features · 24 fixes · 23 contributors
https://t.co/vWJfPSraDx
Use your agent anywhere. 📱💻
Sign in to Cindy on your phone, pick up your desktop context, and send the next task 🚀
Your computer keeps working, and the result syncs back to your phone ✅
Same account, same thread, no lost context.
#Cindy#AI#CrossDevice
DeepSeek cooked with V4 Flash. 🐳
Now live in Cindy — pick it and put it to work 🐈⬛.
Respect to the team for pushing the field forward.
https://t.co/GbfwexoxOr