Claude Haiku 5.5 brings lower prices and an effort setting to Haiku

Anthropic released Claude Haiku 5.5 on October 7, 2026, as the new small model in its Claude 5.5 family. The company positions Haiku 5.5 for high-volume, cost-sensitive work such as summaries, classification and database queries, and as a subagent next to Opus 5.5 and Sonnet 5.5. The same announcement halved Sonnet 5.5 cache read prices and added a monthly API credit for Max and Team subscribers.

What is new in Claude Haiku 5.5

The headline change is price. On the Claude Haiku 5.5 announcement (opens in a new tab), input costs $0.10 and output $0.50 per million tokens for prompts up to 100k tokens. Above 100k tokens, the rates rise to $0.50 input and $2.50 output. Haiku 4.5 is listed at $1.00 input and $5.00 output per million tokens.

Anthropic says Haiku 5.5 costs around 75% less to run than Haiku 4.5 on average. A footnote splits that figure: about 90% cheaper for requests up to 100k tokens, about 50% cheaper above that. The savings estimate already accounts for a newer tokenizer that uses more tokens per task.

Haiku 5.5 is also the first Haiku-class model with an adjustable effort setting. The models overview (opens in a new tab) lists adaptive thinking and a default effort of medium, a 1M token context window, 128K tokens of max output and a reliable knowledge cutoff of Jun 2026.

Benchmarks Anthropic published

Anthropic ran its own comparisons against Haiku 4.5, GPT-6 Luna and Sonnet 5.5. A few rows from that table:

• Terminal-Bench 4.0: 39.2% for Haiku 5.5, against 0.0% for Haiku 4.5 and 70.6% for Sonnet 5.5.
• OSWorld 2.1 (offline subset): 72.4%, against 15.7% for Haiku 4.5 and 83.9% for Sonnet 5.5.
• Humanity's Last Exam without tools: 45.9%, against 10.2% for Haiku 4.5.

These are vendor-reported results. The announcement points to the Haiku 5.5 System Card for evaluation methods.

Who Claude Haiku 5.5 is for

Teams that run large volumes of short, repetitive calls get the most from the price cut. Anthropic's docs describe Haiku 5.5 as built for high-volume, latency-sensitive tasks such as classification, extraction, and routing. Anthropic also names live customer support and browser use, since Haiku 5.5 is its fastest model at standard speed. A footnote adds that the Opus models in Fast Mode are still quicker.

For coding agents, Anthropic pitches Haiku 5.5 as a subagent paired with Opus 5.5 or Sonnet 5.5 on coding work. Users of Claude Code get a built-in path for moving projects across: the migration guide documents a /claude-api migrate command that swaps model IDs and breaking parameters.

How to try Claude Haiku 5.5

The API model ID is claude-haiku-5-5. On Amazon Bedrock it is anthropic.claude-haiku-5-5, and Google Cloud and Microsoft Foundry use claude-haiku-5-5. That ID is fixed, without a date suffix or a separate alias.

The Haiku 5.5 migration guide (opens in a new tab) lists changes that return 400 errors if you only swap the model name:

• Sampling parameters temperature, top_p and top_k must go. A temperature other than 1 fails.
• Assistant prefill is rejected, even with thinking off. End messages with a user turn.
• thinking: {"type": "enabled", "budget_tokens": N} must become {"type": "adaptive"}, with output_config.effort as the control.
• Computer use moves from computer_20250124 to the computer_toolset_20260801 toolset on the Claude API and Google Cloud.

Budget for tokens too. The guide says the same text produces approximately 30% more tokens on Haiku 5.5 than on Haiku 4.5. Thinking blocks from earlier turns now stay in context and count as input. Code that reads the first content block as the answer should select blocks by type.

Other pricing changes in the announcement

Sonnet 5.5 cache reads drop from $0.20 to $0.10 per million tokens. Anthropic estimates this makes Sonnet 5.5 around 20% cheaper on most agentic work.

The new monthly API credit is $100 for Max 5x, $200 for Max 20x and up to $500 for Team, pooled across users. Anthropic says the credit works on any model on the Claude Platform.

What is not known yet

No independent benchmark results for Haiku 5.5 appear on the pages reviewed here. The figures above come from Anthropic's own runs.

The announcement does not say whether Haiku 5.5 is selectable in the Claude apps or on which consumer plans. The migration guide states that Priority Tier is not supported on Haiku 5.5, and that structured outputs are not available for Haiku 5.5 on Amazon Bedrock. Whether either will change is not stated. Haiku 4.5 remains listed as a legacy model, and the docs give no retirement date for it on the overview page.