Replymessage unavailable
deepseek is about 3 months behind because of training on huawei chips and decudaization. Other models like glm-5.1, kimi-k2.6, minimax-m2.7 and small local qwen models are already used by developers and on par with opus for coding. Anthropics only hope remains b2b revenue, which is at risk because once everything settles, companies shift to cost-saving strategies with good enough open models.