Breaking

Anthropic Halves Claude Sonnet 5.5 Cache-Read Price to $0.10 per Million Tokens

October 7, 2026 at 15:28 EDT

Foundation Models
AI Agents
Software Dev & Coding

Anthropic has halved the cache-read price for Claude Sonnet 5.5 to $0.10 per million tokens, saying the change makes the model about 20% cheaper for most agentic work. The cut puts Sonnet's cached-context rate on the same headline level as OpenAI's GPT-6.1 Sol and intensifies the contest around long-running coding agents.

Oct. 7, 2026 | Anthropic

Sonnet 5.5 gets cheaper where agents repeat context

Cache reads drop to $0.10 per million tokens, lowering costs for long-running agent work when prompts reuse cached context.

$0.10

cache read per 1M tokens

50%

cut from the launch read price

~20%

cheaper on most agentic work

token context window

Cache-read price, drawn to scale

Each block represents $0.10 per million cached tokens.

$0.20launch read price

$0.10new read price

The repeated-context portion of an agent run is now half as expensive to read.

How the savings show up

Stable prompt→Cache write→Reused prefix→Lower agent bill

What improves

Cheaper repeated reads for codebases, tool schemas and conversation history.
Better economics for long-running coding and research agents.

What to watch

Savings depend on cache-hit rates, not just the published price.
Cache writes, TTL choices and changing prompt blocks still shape the final bill.

Continue reading

The rest of this article is for AI News Blitz readers. Choose an option below to keep reading.