# Prompt Caching

Anthropic lowered the API list prices for Opus 5.5, the model used in Claude Code: input and output tokens cost 20% less than on Opus 5, and cache reads cost 60% less, with the read rate dropping from one tenth to one twentieth of the input price. What a task costs depends on turns, cache hits, output tokens and model price, because every turn resends the conversation so far. On an illustrative session with identical token counts, the price change alone takes the bill from $3.50 to $2.40, about 31% less; the stated 40% saving for typical workloads also assumes fewer tokens per task at the medium default.

Anthropic lowered the API list prices for Opus 5.5, the model used in Claude Code: input and output tokens cost 20% less than on Opus 5, and cache reads cost 60% less, with the read rate dropping from one tenth to one twentieth of the input price. What a task costs depends on turns, cache hits, output tokens and model price, because every turn resends the conversation so far. On an illustrative session with identical token counts, the price change alone takes the bill from $3.50 to $2.40, about 31% less; the stated 40% saving for typical workloads also assumes fewer tokens per task at the medium default.

In its Claude Platform release notes, Anthropic cut the price of prompt cache reads on Claude Sonnet 5.5 from $0.20 to $0.10 per million tokens, which is 0.05x the base input price instead of 0.1x. Cache writes and all other prices are unchanged. For developers who resend the same long prefix, the saving applies to re-reading the cache, not to the first write or to newly generated content.

In its Claude Platform release notes, Anthropic cut the price of prompt cache reads on Claude Sonnet 5.5 from $0.20 to $0.10 per million tokens, which is 0.05x the base input price instead of 0.1x. Cache writes and all other prices are unchanged. For developers who resend the same long prefix, the saving applies to re-reading the cache, not to the first write or to newly generated content.

Fireworks announced an expanded collaboration with NVIDIA to add capacity in Europe on the NVIDIA Vera Rubin and Grace Blackwell platforms, planning to use the NVIDIA DSX AI factory platform and Spectrum-X Ethernet, with the added capacity supporting European deployments beginning in 2027. The company also introduced enterprise inference capabilities: an Ultrafast inference preview on the NVIDIA Blackwell Platform reaching inference speeds exceeding 1,500 tokens per second, multi-day prompt caching and background mode for long-running agents, and higher rate limits with sub-minute cold starts. These target the latency and cost that arise when a single agent task involves repeated model calls and large amounts of shared context.

Fireworks announced an expanded collaboration with NVIDIA to add capacity in Europe on the NVIDIA Vera Rubin and Grace Blackwell platforms, planning to use the NVIDIA DSX AI factory platform and Spectrum-X Ethernet, with the added capacity supporting European deployments beginning in 2027. The company also introduced enterprise inference capabilities: an Ultrafast inference preview on the NVIDIA Blackwell Platform reaching inference speeds exceeding 1,500 tokens per second, multi-day prompt caching and background mode for long-running agents, and higher rate limits with sub-minute cold starts. These target the latency and cost that arise when a single agent task involves repeated model calls and large amounts of shared context.

Claude Managed Agents on the Claude Platform can now write their own workflows in beta for work with many pieces, such as reviewing hundreds of documents. A workflow is a program that runs many agents in phases and combines their results, and the server runs it in the background as a workflow run. It is enabled by setting the agent's multiagent field to multiagent_20261001 with workflows enabled, and each run is followed through workflow_run.* events on the session's event stream.

Claude Managed Agents on the Claude Platform can now write their own workflows in beta for work with many pieces, such as reviewing hundreds of documents. A workflow is a program that runs many agents in phases and combines their results, and the server runs it in the background as a workflow run. It is enabled by setting the agent's multiagent field to multiagent_20261001 with workflows enabled, and each run is followed through workflow_run.* events on the session's event stream.

Model routing for multi-turn agents is studied at the user-turn boundary, where a router picks one model for the whole turn before it runs. At a matched offloading rate of roughly a fifth of turns, a router that reads session history passes 28.3% of SheetCopilot tasks against 25.0% for the same router seeing only the current request, a 3.3-point gap, while staying within one point of always using the expensive model. A closed-form cache analysis finds downward switches repay their cache cost in about two calls, while upward switches do not break even under the evaluated conditions.

Model routing for multi-turn agents is studied at the user-turn boundary, where a router picks one model for the whole turn before it runs. At a matched offloading rate of roughly a fifth of turns, a router that reads session history passes 28.3% of SheetCopilot tasks against 25.0% for the same router seeing only the current request, a 3.3-point gap, while staying within one point of always using the expensive model. A closed-form cache analysis finds downward switches repay their cache cost in about two calls, while upward switches do not break even under the evaluated conditions.