Ooooh I would love to see third party benchmarks for max. Not just the cherry picked ones by Proton themselves. If their own benchmarks hold up this looks really promising. Also I really wish they shared which open weight model Proton is basing this on. Its not apparent after reading this statement but I really doubt they use an in-house trained model
From the index they show it really looks like for Lumo 2.0 Max they are using GLM-5.2 (max) and Qwen3.5 397B A17B (Reasoning) for Lumo 2.0 Lite.
Ooooh I would love to see third party benchmarks for max. Not just the cherry picked ones by Proton themselves. If their own benchmarks hold up this looks really promising. Also I really wish they shared which open weight model Proton is basing this on. Its not apparent after reading this statement but I really doubt they use an in-house trained model From the index they show it really looks like for Lumo 2.0 Max they are using GLM-5.2 (max) and Qwen3.5 397B A17B (Reasoning) for Lumo 2.0 Lite.