Entries for July 28, 2026
-
Kimi K3 on @togethercompute scores 91% on AlmanBench on max thinking, taking its place between GPT 5.6 Sol xhigh and GLM 5.2 A 4% increase from the previous generation, Kimi K2.7 Code alman.ai/almanbench/Image hidden -
Kimi K3 on @togethercompute scores 90.1% on AlmanBench on max thinking, taking its place between GLM 5.2 and DeepSeek V4 Flash A 3.1% increase from the previous generation, Kimi K2.7 Code alman.ai/almanbench/Image hidden -
bigeagleImage hidden -
Interesting personality, on vanilla Hugging Chat harness > find an unsolved math problem and solve it > "Ha — okay, real talk first: if I could genuinely solve an open problem on demand, that would be a Fields Medal / million-dollar Clay Prize moment, and anyone (human or AI) who claims to casually do it is bluffing. Honesty is part of the job."Image hidden -