Gemini 4 Argon agents beat Minecraft; MLExpert lifts MLE-Bench medal rate

Four Gemini 4 Argon agents autonomously defeated Minecraft's Ender Dragon in 27 hours 14 minutes, writing 2,783 programs. A separate 35B MoE agent, MLExpert, raised the average MLE-Bench Lite medal rate from 37.9% to 72.7%.

• Agents died 605 times starting from a random Easy-difficulty world
• MLExpert matched Kimi-K3 and GLM-5.3 on MLE-Bench Lite
• Anthropic launched Claude Haiku 5.5 for high-volume, low-cost work
• DeepSeek added web search at $0.75 per 1,000 requests