The article details Anthropic’s Claude API pricing as of September 2026, listing per‑million‑token rates for the current model lineup: Haiku 4.5 at $1 input / $5 output, Sonnet 5 at $2 input / $10 output, Opus 5.5 at $4 input / $20 output, and Fable 5.1 at $10 input / $50 output. Cache writes cost 1.25× (5‑min) or 2× (1‑hour) the base input rate, while cache reads are 0.1× (or lower for Opus 5.5 and Fable 5.1). The Batch API halves both input and output prices. All models from Claude 4.6 onward support a 1 million‑token context window at standard rates, with a maximum of 128 k output tokens per request. Because each turn retransmits the full conversation history, input token cost grows super‑linearly; a 40‑turn chat using Sonnet 5 costs $1.09 without caching and $0.29 with a warm 5‑minute cache, scaling to $5.73 and $1.01 for 100 turns. Prompt caching therefore yields the largest savings for repeated prefixes, while the Batch API offers a uniform 50 % discount for asynchronous workloads.