🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Can we design AI Agents that achieve generalizability across diverse task domains? Our new paper introduces OpenHands-Versa, a generalist agent with strong performance on three challenging agent benchmarks, ranking #1 on SWE-Bench Multimodal and The Agent Company leaderboards 🚀

What if we could have *trustworthy* agents that don't just write code, but also do research, understand multimodal content, and perform many practically useful tasks? Today at OpenHands, we released a new agent that gets SOTA or competitive performance on 8 diverse tasks.

I am very excited about both the accuracy results and just my feeling playing around with this new agent. The ability to research things online and comprehend a wide variety of multimodal content has removed many of my blockers in using agents in my everyday work.

"I threw an issue from a side project at 6 AI tools today: Claude Code, OpenAI Codex (the CLI version), Factory, Devin, Jules, and All Hands. None resolved the issue in one-shot, though eventually All Hands got the job done and its PR was merged." yoav.blog/2025/06/10/com…

At All Hands AI we are dedicated to: 1. Making OpenHands the best agent for software development 2. Doing it in an open, community-driven way To make this sustainable, we formed a company around the project. Robert Brennan explains our philosophy in our new blog.

Introducing the OpenHands CLI, a new coding CLI that: - Has top accuracy (similar to Claude Code) - Is completely open source, MIT licensed - Is model agnostic, use an API or bring your own - Is simple to install and run `pip install openhands-ai` and `openhands` (no Docker!)

OpenHands CLI requires Python 3.12 to run - If you run into issues during pip install, you can also use uv (docs.astral.sh/uv/) to launch OpenHands CLI in one line: uvx --python 3.12 --from openhands-ai openhands

Why did the new coding LM Kimi-Dev-72B have a 43% accuracy drop when used in a different harness? The reason lies in the difference between agentic and agentless approaches to doing bug-fixing on repos, as explained in the thread!

For anyone setting up infrastructure for parallel agent rollout (e.g., on SWE-Gym), this Helm chart is exactly what we used for parallel agent rollout and evaluation! And now they are source-available and you can deploy them to your own cluster: github.com/All-Hands-AI/O…