Artificial intelligence (AI) company Anthropic has launched Claude Haiku 5.5, its fastest and most affordable small AI model yet, offering up to 90% lower API prices than its predecessor and matching OpenAI's GPT-6 Luna pricing for shorter requests.
The launch marks the third model in Anthropic's Claude 5.5 family, following Opus 5.5 and Sonnet 5.5, as competition among AI developers increasingly shifts toward performance, speed, and operating costs.
TL;DR
- Anthropic launched Claude Haiku 5.5 on October 7, 2026, targeting high-volume enterprise AI workloads.
- API prices have fallen by 90% for prompts up to 100,000 tokens, with average workload savings estimated at 75%.
- The model matches GPT-6 Luna's base API pricing and outperforms it in several Anthropic-reported benchmarks.
- Haiku 5.5 is available through Anthropic and major cloud platforms.
Anthropic Introduces Claude Haiku 5.5 With Lower API Pricing
According to Anthropic's official announcement, Claude Haiku 5.5 is designed for high-volume, cost-sensitive AI workloads, including document summarization, database queries, information classification, and repetitive enterprise tasks.
The company describes Haiku 5.5 as its fastest and most capable small model, with improvements in knowledge work, computer use, and agentic coding.
However, the most notable development is its pricing.
As reported by VentureBeat, Anthropic has reduced API token prices by 90% for prompts containing up to 100,000 tokens compared with Haiku 4.5.
The updated pricing stands at $0.10 per million input tokens and $0.50 per million output tokens for requests within that threshold, compared with Haiku 4.5's pricing of $1 and $5, respectively.
For prompts exceeding 100,000 tokens, however, Haiku 5.5 costs $0.50 per million input tokens and $2.50 per million output tokens.
This represents a 50% reduction from Haiku 4.5, rather than the headline 90% reduction.
Anthropic says approximately 90% of requests to its previous Haiku model fall within the lower-priced category, while average workloads should cost approximately 75% less to run.
For businesses deploying AI across thousands or millions of requests, these changes could significantly reduce operating expenses, depending on workload sizes and token consumption.
Claude Haiku 5.5 Matches GPT-6 Luna Pricing And Shows Benchmark Gains
The announcement also brings Anthropic into closer pricing competition with OpenAI.
According to OpenAI's official GPT-6 Luna pricing, the model costs $0.10 per million input tokens and $0.50 per million output tokens at its standard lower-context rates.
This matches Claude Haiku 5.5's pricing for prompts up to 100,000 tokens.
However, an important distinction remains. Haiku 5.5 moves to a higher pricing tier above 100,000 tokens, while GPT-6 Luna maintains its base rates until requests exceed 272,000 input tokens.
Beyond pricing, Anthropic also published benchmark results showing substantial performance improvements.
On OSWorld 2.1, which measures how AI agents interact with computers to complete tasks, Haiku 5.5 achieved 72.4% on the offline subset, compared with 15.7% for Haiku 4.5 and 48.9% for GPT-6 Luna.
On Terminal-Bench 4.0, which evaluates agentic coding capabilities, Haiku 5.5 scored 39.2%, compared with 16.4% for GPT-6 Luna. Meanwhile, it recorded a score of 1,620 on GDPval-AA v2.1, surpassing Luna's 1,437.
These figures come from Anthropic's published evaluations and should not be interpreted as independent confirmation that Haiku 5.5 is superior across every workload.
Anthropic also introduced adjustable effort settings for Haiku 5.5, allowing developers to balance model intelligence against speed and operating costs.
Enterprise Customers Report Faster AI Performance
Early enterprise testing has also produced encouraging results, according to customer testimonials included in Anthropic's announcement.
Aaron Vinh, Staff Software Engineer at Asana, said the company observed more than a 30% reduction in task-completion latency when evaluating Haiku 5.5 for its AI Teammates product, alongside inference speeds of up to 2.5 times faster per agent turn.
“We’re very impressed with Claude Haiku 5.5, particularly its speed,” said Vinh.
Meanwhile, HubSpot Distinguished Software Engineer Ze'ev Klapow said Haiku 5.5 achieved a 92.8% average score across three runs of the company's CRM evaluation suite, its strongest result among the models tested.
Yashodha Bhavnani, VP of AI Products at Box, also reported that Haiku 5.5 scored 11 points higher than Haiku 4.5 in early testing while delivering approximately half the latency.
These results indicate potential benefits for customer service automation, business workflows, and AI agents, although they reflect individual organizations' testing environments rather than independently standardized comparisons.
Anthropic Reduces Sonnet 5.5 Costs And Expands Developer Access
Anthropic's announcement extends beyond Haiku 5.5.
The company has also reduced Claude Sonnet 5.5's cache-read pricing by 50%, bringing the cost down from $0.20 to $0.10 per million tokens.
According to Anthropic, this adjustment could make typical agentic workloads approximately 20% cheaper, depending on how frequently applications reuse cached information.
The company is also introducing monthly API credits for Claude Max and Team subscribers.
Max 5x users will receive $100 in monthly credits, while Max 20x subscribers will receive $200. Team subscribers will receive up to $500 in credits shared among users.
These credits are intended to encourage experimentation with AI applications and agents on the Claude Platform.
The launch also introduces stricter cybersecurity safeguards than Haiku 4.5, including restrictions on penetration-testing activities that Anthropic considers more likely to be misused.
AI Pricing Competition Intensifies As Enterprise Adoption Expands
The launch comes as AI developers face increasing pressure to improve model capabilities while lowering the costs of running enterprise applications.
As Yahoo Finance reported, pricing has become an increasingly important consideration for businesses seeking to control AI expenditure.
Topics for more insights:
Companies evaluating AI systems must consider not only model intelligence but also inference expenses, processing speed, and the costs of operating applications at scale.
Anthropic's approach also reflects a growing emphasis on using smaller AI models for routine assignments while reserving more capable models for complex reasoning and coding.
With Claude Haiku 5.5 now matching GPT-6 Luna's base API pricing, businesses have another competitive option for automating repetitive tasks.
However, the pricing advantage will depend on request sizes, output requirements, and real-world performance, making testing against specific business workloads essential.
For Anthropic, the launch strengthens its position in the increasingly competitive market for affordable, high-performance enterprise AI.
