Claude Haiku 5.5 lands at $0.10/$0.50 per million tokens, and Sonnet 5.5 cache reads drop to $0.10

Anthropic released Claude Haiku 5.5 on 7 October 2026, completing the Claude 5.5 family after Opus 5.5 and Sonnet 5.5. The announcement and the model docs went up around 18:00 UTC. The model ID is claude-haiku-5-5. The same announcement halves the price of cache reads on Sonnet 5.5 and adds a monthly API credit to Claude Max and Team plans.

Haiku 5.5 prices

Unlike the rest of the current lineup, Haiku 5.5 is priced by prompt length: a prompt of up to 100,000 tokens pays the lower rates, and a longer prompt pays the higher ones. Per million tokens, from Anthropic’s pricing docs:

Per MTok	Haiku 5.5, prompt up to 100K	Haiku 5.5, prompt over 100K	Haiku 4.5
Input	$0.10	$0.50	$1.00
Output	$0.50	$2.50	$5.00
Cache write (5 min)	$0.125	$0.625	$1.25
Cache write (1 hour)	$0.20	$1.00	$2.00
Cache read	$0.01	$0.05	$0.10
Batch input / output	$0.05 / $0.25	$0.25 / $1.25	$0.50 / $2.50

That is 90% below Haiku 4.5 for short prompts and 50% below it for long ones. Anthropic says about 90% of requests to Haiku 4.5 fell under 100,000 tokens.

Anthropic’s headline figure is smaller: Haiku 5.5 “costs around 75% less to run” on average. The gap is the tokenizer. Haiku 5.5 uses the newer tokenizer from Claude 4.7 onward, so the docs say the same text counts as about 30% more tokens than on Haiku 4.5. Recount prompts before you copy the 90% into a budget.

Sonnet 5.5: one line of the bill

Sonnet 5.5 cache reads drop from $0.20 to $0.10 per million tokens, starting 7 October. Input stays $2, output $10, and 5-minute cache writes $2.50, as on launch day. Anthropic estimates the change makes Sonnet 5.5 about 20% cheaper “on most agentic work”, because cached context is a large share of the tokens an agent loop reads. Workloads with little caching will see almost no change.

The anthropic.com pricing page and the caching section of the pricing docs list $0.10. When we checked, the main model table on the same docs page still showed $0.20 for Sonnet 5.5 cache hits.

What changes for developers

• Context and output: 1M-token context and 128K max output (300K on the Batch API with the output-300k-2026-03-24 beta header).
• Effort: Haiku 5.5 is the first Haiku with adaptive thinking and an effort setting; the API default is medium.
• Breaking changes from Haiku 4.5, per What’s new: manual budget_tokens thinking, non-default temperature/top_p/top_k, and assistant prefill all return errors; computer use needs computer_toolset_20260801 on the Claude API and Google Cloud; responses can start with thinking blocks, so select content by type.
• Retirement: not sooner than 7 October 2027.

Anthropic’s benchmark table is its own. It reports 72.4% on the OSWorld 2.1 offline subset (Haiku 4.5: 15.7%) and 39.2% on Terminal-Bench 4.0, against 70.6% for Sonnet 5.5. Anthropic itself says Sonnet and Opus remain the better choice for complex agentic coding and pitches Haiku at summaries, compaction, classification and subagent work. Artificial Analysis lists the model at 43 on its Intelligence Index at max effort and notes it was very verbose in that run.

Availability

The docs list Haiku 5.5 on the Claude API, Amazon Bedrock (anthropic.claude-haiku-5-5), Google Cloud Vertex AI, Microsoft Foundry and Claude Platform on AWS. Anthropic also published its system prompt for claude.ai and the iOS and Android apps; which Claude plans get it in the model picker is not stated.

Monthly API credits for Max and Team

Rolling out over a few days, per the Help Center: Max 5x gets $100 a month, Max 20x $200, and Team plans $20 per Standard seat and $100 per Premium seat, pooled and capped at $500. Free, Pro and Enterprise are not eligible. You claim the credit by linking one Claude Console organization from claude.ai billing settings. It covers the Claude API, Message Batches, Console Playground, Claude Managed Agents and the Claude Agent SDK on any model. It does not cover interactive Claude Code, extra usage in the apps, or Claude on Bedrock, Vertex AI or Foundry. Unused credit expires each billing cycle.

It replaces the Agent SDK monthly credit announced in June, which the Agent SDK article now says “is no longer available”.