claude-haiku-5-5

claude-haiku-5-5 is Anthropic's efficient-tier model in the Claude 5 family, featuring tiered pricing based on prompt length: $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens, and $0.50 per million input tokens and $2.50 per million output tokens for prompts exceeding 100,000 tokens. Prompt caching reduces repeated input to $0.01/1M (short context) and $0.05/1M (long context). It supports a 1M-token context window, vision inputs, function calling, and streaming. claude-haiku-5-5 is a proprietary model available exclusively through Anthropic's API.

claude-haiku-5-5's tiered pricing structure rewards short-context workloads: at $0.10/$0.50 for prompts under 100K tokens, it is a low-cost option in the Claude 5 family. For prompts over 100K tokens, the rate steps up to $0.50/$2.50 — still competitive for long-context tasks. Prompt caching is available at $0.01/1M (short context) and $0.05/1M (long context). At 1,000 input + 500 output tokens per call (short context), the cost is ($0.10 × 1,000 + $0.50 × 500) ÷ 1,000,000 = $0.00035 per call, or approximately $0.35 per 1,000 API calls — a low per-call cost for high-volume short-context workloads.

claude-haiku-5-5's 1M-token context window is available at both pricing tiers, though the cost step at 100K tokens means long-context workloads are priced differently. Median latency is approximately 200ms TTFT (Estimate). Throughput is approximately 250 tokens per second (Estimate). Maximum output is 32,000 tokens. The short-context pricing tier makes it particularly well-suited for classification, extraction, and summarization tasks where prompts are typically under 100K tokens.

claude-haiku-5-5 is suited to high-volume, cost-sensitive workloads where task complexity is moderate and prompts are typically under 100K tokens: content moderation, entity extraction, document summarization, simple chatbots, code completion for common patterns, and classification pipelines. For workloads with prompts consistently over 100K tokens, the $0.50/$2.50 tier applies — evaluate whether claude-sonnet-5-5 ($2/$10 flat) is more cost-effective for your specific prompt length distribution.

claude-haiku-5-5 occupies the efficient tier in Anthropic's Claude 5 lineup, below claude-sonnet-5-5 ($2/$10) and above the legacy claude-haiku-3 ($0.25/$1.25). Its tiered pricing is unique in the Claude 5 family — no other Claude 5 model uses prompt-length-based pricing. Its intelligence score of approximately 72/100 reflects solid but not frontier capability — adequate for the majority of production NLP tasks at short context lengths. For teams already on Anthropic's API, claude-haiku-5-5 is the natural cost-optimization step for high-volume short-context workloads.