Anthropic's Claude Haiku 5.5 is designed as the fastest and most cost-efficient model in the Claude 5.5 family. For business operators evaluating API expenses, pricing starts at $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens, according to Anthropic's official pricing. Benchmarks published by Artificial Analysis record Haiku 5.5 generating output at 240.5 tokens per second with a context window of 1 million tokens. On Amazon Bedrock, Anthropic highlights customer service as a primary use case, enabling automated systems to resolve service requests quickly.
Claude Haiku 5.5 Pricing and Token Rates
Claude Haiku 5.5 uses a tiered pricing structure based on prompt size:
Prompts up to 100,000 tokens: $0.10 per million input tokens and $0.50 per million output tokens.
Prompts over 100,000 tokens: $0.50 per million input tokens and $2.50 per million output tokens.
Prompt caching pricing on Anthropic's API reflects a 5-minute time-to-live (TTL):
Prompts up to 100,000 tokens: $0.01 per million cache read tokens and $0.125 per million cache write tokens.
Prompts over 100,000 tokens: $0.05 per million cache read tokens and $0.625 per million cache write tokens.
Anthropic also provides additional pricing rules across its model lineup:
Batch processing: Asynchronous batch workloads receive a 50% discount.
US-only inference: Workloads requiring US-only routing carry a 1.1x pricing multiplier on input and output tokens.
The following table compares token rates across the current Claude models and the legacy Haiku 4.5 model listed on Anthropic's pricing documentation:
Model | Context Tier | Input Rate (per MTok) | Output Rate (per MTok) | Prompt Cache Read (per MTok) | Prompt Cache Write (per MTok) |
|---|---|---|---|---|---|
Claude Haiku 5.5 | Prompts up to 100K tokens | $0.10 | $0.50 | $0.01 | $0.125 |
Claude Haiku 5.5 | Prompts over 100K tokens | $0.50 | $2.50 | $0.05 | $0.625 |
Claude Sonnet 5.5 | Standard | $2.00 | $10.00 | $0.10 | $2.50 |
Claude Opus 5.5 | Standard | $4.00 | $20.00 | $0.20 | $5.00 |
Claude Fable 5.1 | Standard | $10.00 | $50.00 | $0.25 | $12.50 |
Claude Haiku 4.5 (Legacy) | Standard | $1.00 | $5.00 | $0.10 | $1.25 |
For prompts that remain within the 100,000-token limit, Haiku 5.5 costs 95% less than Sonnet 5.5 ($0.10 versus $2.00 for inputs, and $0.50 versus $10.00 for outputs). It is also 90% cheaper than the legacy Haiku 4.5 model ($1.00 input and $5.00 output).
Support-Ticket Workload Cost Comparison Across Claude Models
Imagine a small business that processes 10,000 customer support tickets per month using an automated AI triage and response pipeline.
To evaluate API costs, we state the following operational assumptions:
Monthly ticket volume: 10,000 tickets.
Average input per ticket: 1,500 input tokens, encompassing the system prompt, ticket history, and customer inquiry. Because 1,500 tokens is below the 100,000-token boundary, the standard up to 100K rate applies.
Average output per ticket: 500 output tokens for the drafted customer response.
Total monthly input tokens: 10,000 tickets x 1,500 tokens = 15,000,000 input tokens.
Total monthly output tokens: 10,000 tickets x 500 tokens = 5,000,000 output tokens.
Per-million-token costs are calculated using the formula: Cost = (Tokens / 1,000,000) x Rate
Applying this formula to each model produces the following costs:
Claude Haiku 5.5
Input cost: (15,000,000 / 1,000,000) x $0.10 = 15 x $0.10 = $1.50
Output cost: (5,000,000 / 1,000,000) x $0.50 = 5 x $0.50 = $2.50
Total monthly cost: $1.50 + $2.50 = $4.00
Claude Haiku 4.5 (Legacy)
Input cost: (15,000,000 / 1,000,000) x $1.00 = 15 x $1.00 = $15.00
Output cost: (5,000,000 / 1,000,000) x $5.00 = 5 x $5.00 = $25.00
Total monthly cost: $15.00 + $25.00 = $40.00
Claude Sonnet 5.5
Input cost: (15,000,000 / 1,000,000) x $2.00 = 15 x $2.00 = $30.00
Output cost: (5,000,000 / 1,000,000) x $10.00 = 5 x $10.00 = $50.00
Total monthly cost: $30.00 + $50.00 = $80.00
Claude Opus 5.5
Input cost: (15,000,000 / 1,000,000) x $4.00 = 15 x $4.00 = $60.00
Output cost: (5,000,000 / 1,000,000) x $20.00 = 5 x $20.00 = $100.00
Total monthly cost: $60.00 + $100.00 = $160.00
Claude Fable 5.1
Input cost: (15,000,000 / 1,000,000) x $10.00 = 15 x $10.00 = $150.00
Output cost: (5,000,000 / 1,000,000) x $50.00 = 5 x $50.00 = $250.00
Total monthly cost: $150.00 + $250.00 = $400.00
The table below summarizes the total monthly costs and dollar savings across all models for this 10,000-ticket workload:
Model | Monthly Input Tokens | Input Cost | Monthly Output Tokens | Output Cost | Total Monthly Cost | Cost Difference vs. Haiku 5.5 |
|---|---|---|---|---|---|---|
Claude Haiku 5.5 | 15,000,000 | $1.50 | 5,000,000 | $2.50 | $4.00 | Baseline |
Claude Haiku 4.5 | 15,000,000 | $15.00 | 5,000,000 | $25.00 | $40.00 | +$36.00 (10.0x) |
Claude Sonnet 5.5 | 15,000,000 | $30.00 | 5,000,000 | $50.00 | $80.00 | +$76.00 (20.0x) |
Claude Opus 5.5 | 15,000,000 | $60.00 | 5,000,000 | $100.00 | $160.00 | +$156.00 (40.0x) |
Claude Fable 5.1 | 15,000,000 | $150.00 | 5,000,000 | $250.00 | $400.00 | +$396.00 (100.0x) |
Running this support-ticket workload on Claude Haiku 5.5 totals $4.00 per month, delivering a $76.00 monthly savings compared to Sonnet 5.5 ($80.00 per month) and a $36.00 monthly savings compared to Haiku 4.5 ($40.00 per month). Businesses that cache shared support guidelines can further lower prompt costs using prompt caching at $0.01 per million read tokens, or reduce batch ticket processing expenses by 50% with batch processing.
To model how AI token expenses affect your overall software unit economics, calculate your operating figures with our SaaS profit margin calculator.
