Also this week
Sage S2.1 Flash
Sage S2.1 Flash is described as an open-weight science model, and Sage itself reports leading results on the MLE-bench Lite and OfficeQA Pro benchmarks.1 The only coverage here is a brief item in a Latent Space AINews roundup, which gives no scores, rankings or comparison models, and the claim is attributed to Sage rather than independently verified. inference2
The newsletter lists it alongside other model releases, including Liquid d1.2-ultra, which Artificial Analysis ranks #1 on its Decision Index, and the FLUX-mimic-x robotics model, which ranks #1 on RoboLab among real-time models. inference34
What each sentence rests on
Sage S2.1 Flash : The open-weight science model reports leading MLE-bench Lite and OfficeQA Pro results ( Sage).
— Latent Space, [AINews] Perplexity hits a $1B revenue run rate, up 4x in 2026The open-weight science model reports leading MLE-bench Lite and OfficeQA Pro results ( Sage).
— Latent Space, [AINews] Perplexity hits a $1B revenue run rate, up 4x in 2026Liquid d1.2-ultra : It ranks #1 on AA's Decision Index at $0.042 per M input tokens ( AA).
— Latent Space, [AINews] Perplexity hits a $1B revenue run rate, up 4x in 2026FLUX-mimic-x : A 7B video-action robotics model ranks #1 on RoboLab among real-time models ( mimic).
— Latent Space, [AINews] Perplexity hits a $1B revenue run rate, up 4x in 2026