B.AI adds Gemini 3.6 Flash and Gemini 3.5 Flash-Lite to its API with 17% lower token use and 350 tokens/s speeds for high-frequency tasks.

According to @AnthropicAI, the Anthropic Engineering blog published guidance on building more efficient agents that handle more tools while using fewer tokens, featuring code execution with the Model Context Protocol (MCP) (source: @AnthropicAI; Anthropic Engineering blog). The post highlights MCP-based code execution for tool-using agents, indicating standardized context management aimed at token efficiency and scalable multi-tool workflows (source: Anthropic Engineering blog; @AnthropicAI). For trading-focused builders, the documented approach can be applied to AI-driven execution and analysis pipelines to integrate more tools while controlling token usage, though the post does not reference cryptocurrencies or market impacts (source: Anthropic Engineering blog; @AnthropicAI).