💻 Open Models Tack Toward the Frontier

💻 Open Models Tack Toward the Frontier
Tunguz tracks Chatbot Arena rankings from 2023-2026 showing open-weight models like DeepSeek R1 and Kimi K3 repeatedly catch closed frontier systems but never durably lead, as closed models keep producing step-change jumps. This week alone saw Moonshot's Kimi K3 (2.8T parameters), Alibaba's Qwen 3.8 preview and Thinking Machines' Inkling all ship. On price, the median open-weight frontier model runs about 15% cheaper than GPT-5.2-class models, while OpenAI has cut inference costs 50% and Anthropic nears its first profitable quarter — pointing to compressing margins amid one of the largest infrastructure buildouts in US history.
🔗 Read more