Merge(@merge_api):We tested 5 open-weight models against @AnthropicAI Claude Sonnet 5 on 20 real coding tasks. • @Zai_org GLM 5.3 • @Zai_org GLM 5.3 Flash • @deepseek_ai DeepSeek V4 Pro • @deepseek_ai DeepSeek V4 Flash • @Kimi_Moonshot Kimi K3 GLM 5.3 won at a 1/10 of Claude's cost:

2026.09.15 13:30

We tested 5 open-weight models against @AnthropicAI Claude Sonnet 5 on 20 real coding tasks. • @Zai_org GLM 5.3 • @Zai_org GLM 5.3 Flash • @deepseek_ai DeepSeek V4 Pro • @deepseek_ai DeepSeek V4 Flash • @Kimi_Moonshot Kimi K3 GLM 5.3 won at a 1/10 of Claude's cost: