🚀 Thrilled to announce Dream-Coder 7B — the most powerful open diffusion code  LLM to date.

Wrapped up a SWE-Perf website redesign using Qwen3-Coder on AnyCoder (huggingface.co/spaces/akhaliq…). The process was incredibly fast and great! One question for Qwen devs, though: did you pretrain a secret love for the color purple into the coder's persona? 😉

👋 IMO OpenHands will become the default "inference engine" for coding agents

Qwen3-Coder-Flash (size == 30B-A3B) just landed. Qwen Code (0.0.1-alpha.12) also picked up a few upgrades. Tiny ≠ trivial, and lightweight ≠ light-headed. Will keep iterating and aim for better agentic coding.

I feel like this completely flew under the radar despite being a huge deal for discrete diffusion models: DremOn is a 7B dLLM that can do variable length generation, solving something that has been a huge challenge! The idea is clever: Let's just randomly insert <|delete|>

Log in to Qwen Code with your Qwen Chat account and get 2000 free daily requests for Qwen3 Coder 480B!

Dimitri von Rütte Thanks for highlighting this! Very clever approach, and the blog post is great! hkunlp.github.io/blog/2025/drea…

Token crisis: solved. ✅ We pre-trained diffusion language models (DLMs) vs. autoregressive (AR) models from scratch — up to 8B params, 480B tokens, 480 epochs. Findings: > DLMs beat AR when tokens are limited, with >3× data potential. > A 1B DLM trained on just 1B tokens

Excited to see Qwen3-Coder 480B as the default model for AK’s anycoder — thanks! Gave it a one-shot prompt to build an interactive Win95 desktop, and it just works!

1–2 years ago, when I first started training text diffusion models, I had this empirical feeling that they could handle more epochs of training data. It’s great to now see the community sharing experiment logs using the "LM as physics" research approach.🤗

To be fair, I’m not saying there is no hope - it’s just that there is no evidence that the cross point exists in the non-overfitting regime.

What’s particularly striking is that 1B unique tokens trained for 96 epochs can match the performance of 96B unique tokens trained for a single epoch. At first glance, this seems counterintuitive. However, if we randomly mask tokens during training, a sequence of length L can

GPT-5, think more. In our latest LiveCodeBench Pro tests for Competitive Programming, GPT-5 Thinking hit a true 0→1 moment in 2025 Q1 set, the only model to crack the hard split, and this wasn’t even GPT-5 Thinking Pro. Average response length exceeded 100,000 tokens, which is

Where are our computer‑use agents (CUA) standing on OSWorld‑Verified? Potentially already ~80%. We made this analysis, which summarizes the latest OSWorld-Verified submissions with 27 models evaluated over 369 tasks, and conducted a case study on the o3+Jedi-7B approach to

We are super excited to release OpenCUA — the first from 0 to 1 computer-use agent foundation model framework and open-source SOTA model OpenCUA-32B, matching top proprietary models on OSWorld-Verified, with full infrastructure and data. 🔗 [Paper] arxiv.org/abs/2508.09123 📌

Someone asked me: "MCP is so hot right now. If the human-computer interface completely changes in the future to just a chat box, wouldn't your computer use work become useless?" I said: "Maybe, that's possible in one timeline. But it's also possible things will be different

As computer-use agents (CUAs) handle critical digital tasks, open research is key to study their capabilities, risks. 🚀After a year, we release OpenCUA: 1) largest CUA dataset/tool, 2) training recipe, 3) ~SOTA model on OSWorld. Released to drive transparent,safe CUA research!

Check out our latest open-source project, OpenCUA, for Computer Use Agents (CUAs)! Find the code, annotation tool, the largest CUA dataset, scalable training recipe, and state-of-the-art model on OSWorld at: opencua.xlang.ai.

Qwen Code began as a fun side project with Fan Zhou and Binyuan Hui to explore Qwen3 Coder’s terminal-based agentic skills. Right before launch, we decided to share it as a preview with free quota alongside Qwen3 Coder, so the community could test its agent skills in the wild.

Thanks for sharing our work😄