Hacker News

New stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Google Beam expands to new countries(blog.google)
    discuss
  2. Brazil's Lula Campaign Dataset(huggingface.co)
    discuss
  3. Starbucks Korea CEO dismissed, what about startups?(2nd-look.vercel.app)
    discuss
  4. GitCaliber, swap github.com for gitcaliber.com to get a Git repo explained(gitcaliber.com)
    discuss
  5. Show HN: I made an app that transform your daily habits into a medieval(reddit.com)
    discuss
  6. City council can approve license plate cameras without you ever finding out(alprradar.com)
    discuss
  7. My relationship with AI: seductive, frustrating and always demanding vigilance(theguardian.com)
    discuss
  8. Dr Tricks Senior Dev N2 Taking Job W AI
    discuss
  9. If Carpenters Were Hired Like Programmers (2013)(archive.org)
    discuss
  10. My Story(github.com/sovereign-intel-founder)
    discuss
  11. The Unhired Index: Big Tech revenue-headcount trend suggests 185K missing hires(drorpoleg.com)
    2comments
  12. Security Debt: How Legacy Systems Increase Cyber Risk(cybrsecmedia.com)
    discuss
  13. How Apple uses faulty silicon to power fab new devices(cultofmac.com)
    discuss
  14. AOS GLM Language Engine – Beta Test(github.com/thepikey)
    discuss
  15. Ask HN: How do you test Lambda/ECS code before deploying?
    2comments
  16. Work, Wellbeing, and Choice: Empirical Lessons for AI Futures(arxiv.org)
    discuss
  17. Local AI News Feed(local-ai-signal.vercel.app)
    2comments
  18. The Fine-Grained City Needs an Operating System(courtyardurbanist.com)
    discuss
  19. Small Modular Reactors: a short history of a nuclear pipe dream(engelsbergideas.com)
    discuss
  20. LensVLM: Compressing long context as images, expanding only relevant pages(huggingface.co)
    discuss
  21. Ask HN: Do you offload your coding to AI, or keep your skills sharp?
    3comments
  22. Can internal model transparency tame the AI race?(karthiktadepalli.com)
    discuss
  23. The meta-harness for coding agents(soloterm.com)
    discuss
  24. Show HN: Jev-mice – mouse colony simulation using Jev and deterministic engine(carsonsweet.com)
    discuss
  25. Discover Solo: The Ultimate AI-Integrated Control Panel(master.dev)
    discuss
  26. Steam Frame Teardown(ifixit.com)
    discuss
  27. SpaceX Starshield(wikipedia.org)
    discuss
  28. Mexican Navy's tall ship Cuauhtémoc will be at the Embarcadero all week(kqed.org)
    discuss
  29. A Worthwhile Trade(nejm.org)
    discuss
  30. Loopjacking in A2A Implementations: Hijacking Human-in-the-Loop Approvals(adithyanak.com)
    discuss

Qwen Image 2.1 beats Google Nano Banana 2.0 with minuscule 7B parameter model

4 pointsby 1h agotomshardware.com
1 comments
1h agoHN ↗

I haven't done enough tests on the editing capabilities but as far as the strict text to image goes, Alibaba is reaching to claim it is comparable to NB 2.

It's definitely a nice upgrade from the last open-weight version (Qwen-Image 1.0) at least in terms of strict coherence and prompt understanding... but a lot of the outputs seem distilled for lack of a better word - likely trained on poor synthetic data. There's also some elements of tinging that very much reminds me of early gpt-image outputs.

You can mitigate it a bit using better samplers (like res_2m paired with the beta_57 scheduler, etc). A lot of people have also seen better results using a higher CFG than what is recommended by the Qwen team.

On my GenAI Image Showdown bench which emphasizes adherence to prompts, Qwen-Image 2.1 clocked in at 7 out of 15 which is SOTA for an open-weight model (Ideogram 4 is the only other open-weight model that outscored it), but the quality is frustratingly inconsistent.

https://genai-showdown.specr.net/?models=local