Archive · 8 Oct 2026
Anthropic Launches Haiku 5.5 With Lower Costs and Effort Controls
Anthropic has released a new AI model called Claude Haiku 5.5.
The company says it is faster, more capable and cheaper to run than its earlier Haiku 4.5 model.
It is designed for many repeated jobs, such as sorting information, summarizing documents and helping customers.
It can also handle smaller jobs for Anthropic’s larger AI models.
Developers can choose whether it should use fewer resources for quick, simple tasks or work harder when needed.
Anthropic says it did better than the older model in some tests, though those results are company-reported.
Its safety rules block penetration testing but allow some defensive cybersecurity work.
The model is available through Anthropic and several major cloud services.
Anthropic says Claude Haiku 5.5 is faster and more capable than Haiku 4.5, while costing about 75% less to run on average.
For prompts up to 100,000 tokens, input costs $0.10 and output costs $0.50 per million tokens; longer prompts cost $0.50 and $2.50, respectively.
The model is aimed at high-volume tasks such as summarization, classification, extraction, database queries, customer support and voice agents.
Its adjustable effort setting lets developers trade off speed and cost against greater computational effort; it can also serve as a subagent alongside Opus 5.5 and Sonnet 5.5.
Anthropic says the model has improved alignment results and cybersecurity safeguards that block penetration testing; it is available through Anthropic and major cloud platforms.
- Who
- Anthropic launched Claude Haiku 5.5.
- What
- A small AI model focused on speed and lower operating costs, with adjustable effort settings and updated safeguards.
- When
- Wednesday; the articles do not specify a calendar date.
- Where
- Anthropic’s platforms, Amazon Web Services, Google Cloud and Microsoft Azure.
- Why
- Anthropic positions it for fast, high-volume, cost-sensitive tasks and as a subagent alongside larger models.
Reported benefits
Limits and qualifications
Performance and intended use
Reported benefits
Anthropic describes Haiku 5.5 as its fastest and most capable small model, suited to quick, high-volume tasks and work as a subagent.
Limits and qualifications
Anthropic says Sonnet 5.5 and Opus 5.5 remain better choices for complex agentic coding.
Benchmark claims
Reported benefits
Anthropic reports improved results over Haiku 4.5 on OSWorld and Terminal-Bench; customer examples also report latency and speed gains.
Limits and qualifications
The benchmark figures are company-reported, and the articles do not provide independent verification.
Cybersecurity safeguards
Reported benefits
The safeguards permit some defensive cybersecurity work and, according to one article, a broader range of defensive tasks than Sonnet 5.5 allows.
Limits and qualifications
They still block penetration testing and techniques Anthropic considers more likely to be used by attackers.
- Average operating-cost reduction
- Anthropic says Haiku 5.5 costs about 75% less to run on average than Haiku 4.5.
- Cost for prompts up to 100,000 tokens
- Anthropic says costs are 90% lower than Haiku 4.5; this prompt length accounted for around 90% of requests to the earlier model.
- Cost for longer prompts
- Anthropic says costs are 50% lower than Haiku 4.5 for prompts over 100,000 tokens.
- Pricing up to 100,000 tokens
- $0.10 per million input tokens and $0.50 per million output tokens.
- Pricing over 100,000 tokens
- $0.50 per million input tokens and $2.50 per million output tokens.
- Availability
- Anthropic platforms, Amazon Web Services, Google Cloud and Microsoft Azure.
- Sonnet 5.5 cache reads
- Anthropic cut the price to $0.10 per million tokens and says this can make most agentic work around 20% cheaper.
Sources
Anthropic launches Claude Haiku 5.5, claims it is 75% cheaper than previous model
Anthropic Launches Claude Haiku 5.5 with faster performance and new effort controls
Anthropic Launches Claude Haiku 5.5, Cutting The Cost Of Its Smallest Model By Around 75%: All Details
