Anthropic Releases Claude Haiku 5.5: Fastest AI Model with 75% Cost Reduction

Reviewed byNidhi Govil

10 Sources

Share

Anthropic launched Claude Haiku 5.5, cutting API token prices by 90% for requests under 100,000 tokens. Starting at $0.10 per million input tokens, the new AI model matches OpenAI's GPT-6 Luna pricing while delivering superior benchmark performance. The release targets high-volume cost-sensitive tasks like live customer support and document summarization.

News article

Anthropic Cuts AI Model Costs with Claude Haiku 5.5 Launch

Anthropic released Claude Haiku 5.5, positioning it as the fastest and most affordable model in its lineup

1

. The AI model targets high-volume cost-sensitive tasks including document summarization, database queries, and live customer support

2

. Starting at $0.10 per million input tokens and $0.50 per million output tokens, Haiku 5.5 delivers a 90% cost reduction for requests below 100,000 tokens compared to its predecessor

3

. Anthropic estimates workloads cost approximately 75% less to run than on Haiku 4.5, accounting for request sizes and token consumption changes

3

.

Matching OpenAI's GPT-6 Luna Pricing While Exceeding Performance

The new pricing structure matches OpenAI's GPT-6 Luna rates exactly, creating direct competition in the budget AI model segment

4

. However, benchmark results show Claude Haiku 5.5 outperforming Luna across all tests Anthropic shared

1

. On OSWorld 2.1, which tests whether AI can operate real computers through multi-step tasks, Haiku 5.5 scored 72.4% versus Luna's 48.9%

4

. Terminal-Bench 4.0 results were even more striking, with Haiku 5.5 achieving 39.2% compared to Luna's 16.4% and Haiku 4.5's 0%

4

. For context, Sonnet 5.5 scored 70.6% on the same test

4

.

Tiered Pricing Structure and Adjustable Effort Settings

The aggressive cost reduction applies specifically to shorter requests. Above 100,000 tokens, Haiku 5.5 costs $0.50 per million input tokens and $2.50 per million output tokens, representing a 50% reduction from Haiku 4.5 rather than 90%

3

. Anthropic reports approximately 90% of Haiku 4.5 requests fell into the shorter category

3

. Claude Haiku 5.5 introduces adjustable effort settings for the first time in the Haiku line, allowing users to select Low, Medium, High, Xhigh, or Max settings depending on task complexity

1

. Higher effort settings produce more intelligent responses but increase costs proportionally

1

.

Strategic Role in Enterprise Agentic Coding Workflows

Anthropic positions Haiku 5.5 as a supporting worker for Opus 5.5 and Sonnet 5.5 in agentic coding tasks

3

. A larger model might assemble a financial presentation while Haiku retrieves specific data points like revenue figures for individual slides, according to Alex Wang from financial AI company Rogo

3

. "It's accurate enough that we'd trust it there and fast and cheap enough that we can run it a lot," Wang stated

3

. The model pairs well with more capable siblings for complex projects requiring both sophisticated reasoning and rapid execution of routine subtasks

2

.

Customer Validation Shows Significant Speed Improvements

Asana tested Haiku 5.5 before release, running it through evaluation suites for its AI Teammates agent

5

. Task completion latency dropped more than 30%, while inference per agent turn accelerated by as much as 2.5 times compared to their current model

5

. "It's a noticeably snappier experience," said Aaron Vinh, staff software engineer at Asana

5

. HubSpot's Ze'ev Klapow reported a 92.8% average across three runs of its CRM evaluation, the strongest result among models tested

3

. Box's VP of AI Products, Yashodha Bhavnani, noted an 11-point improvement over Haiku 4.5 with approximately half the latency

3

.

Additional Cost Reductions and API Credits for Subscribers

Beyond the Haiku launch, Anthropic halved Sonnet 5.5 cache read prices from $0.20 to $0.10 per million tokens

5

. The company estimates approximately 20% savings on typical agent workloads, though actual savings depend on how much stored context applications reuse

3

. Monthly API credits roll out this week: $100 for Max 5x subscribers, $200 for Max 20x, and up to $500 shared among Team subscription users

4

. These credits can be spent on any model through the Claude Platform

3

. The model is available now on the Claude Platform, AWS, Google Cloud, and Microsoft Azure under the name claude-haiku-5-5

4

.

Today's Top Stories

© 2026 TheOutpost.AI All rights reserved