Users are praising Upstage's Solar-Open2 250B open model release for its strong SWE-Bench scores matching bigger closed models and the major jump in open weights scale.
No Digg Deeper questions have been answered for this story yet.
Most Activity
Solar Open2 250B just dropped on Hugging Face https://huggingface.co/upstage/Solar-Open2-250B
Super happy to announce @upstageai' new model, #SolarOpen2. It's a very good model. Please try it out: https://huggingface.co/upstage/Solar-Open2-250B
@_akhaliq Incredible every 6 hours a new model is released. I am to slow as a human to follow this pace
@hunkims @upstageai Another big open model enters the race👀 250B is serious.
@hunkims @upstageai Congrats! Did you really only train with NoPe on _all_ layers?
@hunkims @upstageai Adorei os resultados, mas tenho uma dúvida. Porque razão foram comparados com modelos tão ultrapassados como os da Mistral mais antigos?
@_akhaliq 70.4 On SWE Bench Verified at 250B is nuts, that's basically matching much bigger closed models. The MCP Atlas jump over Solar 100B is the one that surprised me most though.
@_akhaliq 250B is a wild jump from the last release makes you wonder what they changed under the hood
@hunkims @upstageai Great release ! Congrats to the team !
@hunkims @upstageai Looks like another capable model
@_akhaliq Man 250B is a whole meal, my GPU is sweating just looking at it 😅
@_akhaliq open weights matching proprietary reasoning benchmarks changes deployment expenses completely. local hosting finally makes economic sense for production apps.
@_akhaliq Open-weight models at this scale were unthinkable a year ago Now they just appear on Hugging Face like any other release
@_akhaliq Another day, another model making my bookmarks folder heavier. 😂📚
@_akhaliq wow thats great
@_akhaliq Finally, something to power all the models I'll be training to predict your next tweet
@hunkims @upstageai Great to see new models from Korea!