deepseek-v4.1IMAGE
DeepSeek V4.1 Flash on an induction benchmark: beats older DeepSeek, sits between GPT-5.6 and GPT-6 Luna
A custom induction benchmark finds V4.1 Flash-1017 clearly better than other DeepSeek models, landing between GPT-5.6 Luna and GPT-6 Luna — decent but not impressive.
Serafim Batzoglou
@s_batzoglouArchived 12 days agoOriginally posted 2026.10.18
Serafim Batzoglou @s_batzoglouOriginally posted 2026.10.18
DeepSeek v4.1 Flash-1017 is a good model for its very low price and open weights. I benchmarked it on induction, and it does significantly better than all the other DeepSeek models. Performance between GPT-5.6 Luna and GPT-6 Luna. Not bad, and not impressive either. GitHub: https://github.com/SerafimBatzoglou/concept-synth/blob/main/benchmarks/induction/docs/leaderboard.md
x.com/s_batzoglou/status/2111679756924260670Archived 2026.10.18 16:24