Pricing Update
Anthropic
-
New model added:
Claude Haiku 5.5. -
New pricing tier for Claude Haiku 5.5 based on prompt length:
- Up to 100,000 tokens: input $0.10/MTok, 5m cache write $0.125/MTok, 1h cache write $0.20/MTok, cache hit $0.01/MTok, output $0.50/MTok
- Over 100,000 tokens: input $0.50/MTok, 5m cache write $0.625/MTok, 1h cache write $1/MTok, cache hit $0.05/MTok, output $2.50/MTok
-
Claude Sonnet 5.5 cache-hit pricing changed:
- was $0.20/MTok
- now $0.10/MTok
- Documentation now states Sonnet 5.5 cache hits use a 0.05x multiplier instead of the standard 0.1x.
-
Batch pricing added for Claude Haiku 5.5, also tiered by prompt length:
- Up to 100,000 tokens: input $0.05/MTok, output $0.25/MTok
- Over 100,000 tokens: input $0.25/MTok, output $1.25/MTok
-
Long-context billing rule changed:
- Previously, all Claude 4.6+ models and Claude Mythos Preview had the full 1M context window at standard pricing.
- Now Claude Haiku 5.5 is explicitly excluded from that standard-pricing rule and instead uses higher prices for prompts over 100,000 tokens.
-
Tool use pricing table updated for Claude Haiku 5.5:
- added tool-use system prompt token overhead:
- 286 tokens for
tool_choice: auto/none - 406 tokens for
tool_choice: any/tool
- 286 tokens for
- added tool-use system prompt token overhead:
-
Retirement availability changed for Claude Haiku 3.5:
- was retired except on Bedrock and Google Cloud
- now retired except on Google Cloud only.