Build Intelligent, Useful and Trustworthy Robots to Make Our Life Better
Joined September 2025
- Tweets163
- Following24
- Followers772
- Likes63
Meet your new bartender. 🍸
Powered by Dexmal's DM0.5, this robot grasps cups of different sizes, adapts to different shapes, and precisely pours liquids into various containers. What looks effortless takes careful perception and precise motion control.
Cheers to embodied AI!
Here’s Dexmal’s DM0.5 folding cardboard boxes autonomously, using visual input and robot state to generate continuous actions. It coordinates both arms as the cardboard changes shape and contact conditions change.
We’ve run the same task on both W1 and ARX, demonstrating bimanual control across different robot platforms.
Plug & play? More like plug & precision. 🔌
With Dexmal's DM0.5, robots autonomously plug, unplug, and repeat with sub-0.5mm alignment precision, bringing greater reliability to electronics assembly and connector testing.
Dexbotic 2.0 Feature Spotlight: More Ways to Evaluate VLAs 📊
Dexbotic 2.0 now supports two additional benchmarks, VLA-Arena and RoboDojo.
VLA-Arena tests safety, robustness, generalization, and long-horizon behavior across L0–L2 difficulty levels. RoboDojo tests VLA generalization on complex bimanual manipulation tasks.
With these additions, Dexbotic 2.0 now supports a broader range of widely used VLA benchmarks.
Open sourced: github.com/dexmal/dexbotic
Dexbotic 2.0 Feature Spotlight: Observation History 🕒
In a multi-step task, the current camera view may not show what the robot has already done. To decide what comes next, the model needs context from earlier in the task.
Dexbotic 2.0 now supports observation history for DM0.5, helping it track progress through longer tasks.
Open sourced: github.com/dexmal/dexbotic
Dexbotic 2.0 Feature Spotlight: Faster DM0.5 Inference ⚡
Inference speed is the real test of a VLA in deployment.
Dexbotic 2.0's new optional inference component makes DM0.5 up to ~5× faster, with no change to the model, accuracy or hardware.
Your call: go fast, or stay on the default.
Open sourced: github.com/dexmal/dexbotic
Understanding the structure and behavior of the physical world is a big goal for AI. For robotics, that understanding must also translate into action.
At Dexmal, we’re working across world models, robot policies, training, inference, and real-world execution—connecting each step so robots can learn and act in the physical world.
A thoughtful perspective from @drfeifei on where this field is headed. Congratulations to World Labs and AMD on this next chapter!
A different object on the table shouldn’t become another task on a developer’s to-do list.
Dexmal DM0.5 uses the same model weights to recognize, locate, and grasp objects across different shapes, materials, and positions—even with external disruptions. No model swaps or parameter tweaks required.
That's less repetitive engineering for developers, and faster adaptation to whatever the next scenario throws at it.
Watch DM0.5 in action 👇
Memory is one of the hardest problems in embodied intelligence.
A robot doesn't just need to see what's in front of it. It needs to remember what happened before, especially across long, multi-step tasks.
Dexmal DM0.5 carries native memory of up to 60 seconds, enough to hold the full arc of a multi-step task without losing track.
In this desktop-cleaning demo, the robot remembers exactly where each object came from. Every item goes back to its original spot, in the right order.
Back in July, we gave Dexmal a complete brand reset.
A new logo, a new visual system, and a clearer way to express what we’re building toward.
The idea underneath it never changed. Build better models, build systems that actually work, and build trust by testing them in the real world.
This was the film we made to tell that story.
Intelligent. Useful. Trustworthy.
GPT-6 Astra is genuinely impressive.
It's sparking a debate: if a general-purpose model can see, reason, and pilot a robot arm, does embodied AI still need its own foundation model?
We think it still does. Astra can do this because it was trained on robot data — that's not proof physical intelligence emerges from language and vision alone. Robotics doesn't get the shortcuts coding did. It still needs real robot data, real-world evaluation, and contact-rich physical interaction that no amount of general scale replaces.
General intelligence is getting better at understanding the physical world. Embodied intelligence is about learning how to act in it.
Your embodied model works in the lab. Can developers actually run it?
DexDev MaaS makes general-purpose embodied models as accessible as any cloud API.
A single instruction is enough to start — the model perceives its environment, understands the task, and acts. Teams working with their own data, whether in LeRobot v3 format or otherwise, can take the full post-training workflow online, from upload to fine-tuning to deployment.
Inference, fine-tuning, and deployment all happening in one place.
maas.dexmal.com
Most world models generate plausible futures. DW0.5 judges them.
Every prediction is conditioned on the action that caused it. Push a cup left or right, and the outcome should diverge accordingly — then a Value Expert judges which future is actually worth pursuing.
That's what makes DW0.5 a training ground, not a demo reel. It simulates success, failure, drift, and recovery, all of which trained on real teleop, first-person, and multi-source data.
It's also the first world model to close this loop end-to-end, from simulation straight into post-training.
A 60% reduction in real-world training data. A 40% drop in training costs. Now #1 on WorldArena and EWMBench.
DW0.5 is open-sourced on GitHub and Hugging Face.
github.com/dexmal/opendw
