#agentic benchmarks

BottleCap AI Releases ThinkingCap-Qwen3.8-27B: 37.2% FewerNEW⭐8

ThinkingCap-Qwen3.8-27B by BottleCap AI cuts thinking tokens by 37.2% on average across 12 benchmarks, with only a 0.86pp accuracy drop.