What Anthropic Shipped
Claude Haiku 5.5 is the third model in the Claude 5.5 family after Opus 5.5 and Sonnet 5.5, released October 7, 2026. It is Anthropic's smallest, fastest tier — built for high-volume jobs like summaries, document extraction, classification, and subagent tasks. It is live on the Claude Platform, Amazon Bedrock, Google Cloud, and Microsoft Azure.
The Pricing Split Everyone Should Understand
Pricing splits at 100K prompt tokens: up to that line, input costs $0.10 and output $0.50 per million tokens — about 90% below Haiku 4.5's $1/$5. Above 100K, rates jump to $0.50/$2.50, still a 50% cut. Anthropic estimates it runs about 75% cheaper than Haiku 4.5 on average, and the Batch API takes another 50% off. The short-context price exactly matches OpenAI's GPT-6 Luna.
| Haiku 5.5 pricing per million tokens | Details |
|---|---|
| Prompts up to 100K | Input: $0.10, Output: $0.50 |
| Prompts over 100K | Input: $0.50, Output: $2.50 |
| Haiku 4.5 (old) | Input: $1.00, Output: $5.00 |
| GPT-6 Luna (short) | Input: $0.10, Output: $0.50 |
Bigger Window, New Effort Dial
Haiku 5.5 carries a one-million-token context window (up from 200K on Haiku 4.5) and up to 128K output tokens. It is the first Haiku-class model with an adjustable effort setting — adaptive thinking is on by default, and the effort parameter (default medium) lets a developer dial the same request toward lower cost or higher intelligence instead of switching model tiers.
Benchmarks: Where It Beats GPT-6 Luna
Anthropic's published figures put Haiku 5.5 ahead of GPT-6 Luna on every shared benchmark: OSWorld 2.1 desktop operation 72.4% vs 48.9%, Terminal-Bench 4.0 39.2% vs 16.4%, FrontierCode 46.4% vs 42.4%, and GDPval-AA office work 1620 vs 1437. These are vendor-reported numbers, but independent Artificial Analysis data also puts Haiku 5.5 at the top of its price class on the Intelligence Index at maximum effort.
Frequently Asked Questions
What is Claude Haiku 5.5?
It is Anthropic's smallest and fastest model, released October 7, 2026, with a 1M-token context window, up to 128K output tokens, adaptive thinking, and the first adjustable effort setting on a Haiku model. It targets high-volume work like summaries, extraction, and subagent tasks.
How much does Claude Haiku 5.5 cost?
$0.10 input / $0.50 output per million tokens for prompts up to 100K tokens, then $0.50 / $2.50 above 100K. That is about 90% below Haiku 4.5 at short contexts and roughly 75% cheaper on average.
How does Haiku 5.5 compare to GPT-6 Luna?
List prices match at short contexts, but Anthropic's benchmarks show Haiku 5.5 ahead of Luna on desktop operation (OSWorld 72.4% vs 48.9%), terminal work (39.2% vs 16.4%), and office tasks (GDPval-AA 1620 vs 1437). Sonnet 5.5 still leads on complex agentic coding.
Where is Haiku 5.5 available?
On the Claude Platform as claude-haiku-5-5, plus Amazon Bedrock, Google Cloud, and Microsoft Azure. Batch jobs support up to 300K output tokens in beta at 50% off.
