According to @DeepLearningAI, Anthropic launched Claude Sonnet 4.5 with a variable reasoning-token budget, expanded 200k–1M token context windows, improved coding and reasoning benchmark performance, free web access for trials, and API pricing set at $3 per million input tokens and $15 per million output tokens, which directly affects cost planning for AI-assisted research and trading pipelines; Source: @DeepLearningAI on X, Oct 14, 2025. According to @DeepLearningAI, Anthropic also released a Claude Agent SDK and refreshed Claude Code, adding automatic context tracking and summarization, a persistent memory tool, checkpoints with safe rollbacks, and a VS Code–compatible IDE extension that can support production agent workflows for code and data tasks used by quant and crypto teams; Source: @DeepLearningAI on X, Oct 14, 2025. According to @DeepLearningAI, based on the stated API rates, an input of 200,000 tokens would cost about $0.60 and 1,000,000 input tokens about $3.00, while 10,000 output tokens would cost about $0.15, providing concrete per-task budgeting parameters for AI-driven trading research and code analysis; Source: @DeepLearningAI on X, Oct 14, 2025.