Anthropic Claude Haiku 5.5
Anthropic's fastest model, built for high-volume, latency-sensitive work such as classification, extraction, routing and subagent tasks, with vision, tool use and a 1M-token context window.
Anthropic's Claude Haiku 5.5 is a cloud text model. It is multimodal: alongside a text prompt it accepts images, then replies with generated text. Its context window handles up to 1M input tokens and up to 128K output tokens, and it supports adjustable reasoning effort for harder problems. Generation can be tuned with system instructions. It runs through OpenRouter and Anthropic using your own API key, from $0.10 per million input tokens.
- Released
- Oct 2026
- Pricing
- $0.10 / 1M in · $0.50 / 1M out
- Inputs
- Text, Images
- Context window
- 1M in · 128K out
- Reasoning
- Adjustable effort
- Controls
- System prompt
Examples
Generated with Claude Haiku 5.5 via OpenRouter. The same three prompts run against every text model in the catalog, shown verbatim, so the only thing that changes between two models’ answers is the model.
Explain what an API is to someone who has never written code. Do not use a restaurant or waiter analogy. Plain prose only — no headings, bullet points, or markdown formatting. Maximum 50 words.
An API is a set of rules that lets one piece of software ask another for information or services in a standard way. Think of it as a published list of requests a program understands, along with the format of the answers it will send back. Your weather app, for example, uses one to fetch forecasts from a weather service.
A two-person startup is deciding whether to build their own user authentication or use a hosted provider. Give them a clear recommendation, then the single strongest argument against your own recommendation. Plain prose only — no headings, bullet points, or markdown formatting. Maximum 150 words.
My recommendation is to use a hosted authentication provider such as Auth0, Clerk, or Supabase Auth. Authentication is a security-critical problem where subtle mistakes, like flawed password reset flows, session handling bugs, or weak credential storage, can expose every user at once. Two people cannot realistically match the attention that specialized providers give to these areas, and building it yourselves diverts time from the product that actually differentiates your startup. Hosted providers also handle compliance features, multi-factor authentication, and social login out of the box, and you can migrate later if needs change. The strongest argument against this recommendation is cost and lock-in. Per-user pricing can climb steeply as the startup grows, sometimes becoming a significant line item before revenue justifies it. Moving off a provider later means migrating password hashes, sessions, and user identifiers, which vendors often make difficult, so early dependence can quietly constrain future decisions.
Write a briefing for a non-technical executive explaining why their company's spending on AI is hard to predict, and what they can do about it. Cover at least three distinct causes. Plain prose only — no headings, bullet points, or markdown formatting. Write exactly 205 words.
Our AI spending is difficult to forecast for several connected reasons, and understanding them is the first step toward controlling it. Most AI services charge by usage rather than by a fixed subscription, so every question, document, or generated report carries a small cost that multiplies with volume. A single employee experiment can look harmless, while an automated workflow that runs thousands of times overnight can quietly produce a large bill. Usage also swings with business cycles, seasonal demand, and the unpredictable ways staff discover new applications. A second cause is that the price of the underlying technology keeps changing. Providers regularly release new models, retire old ones, and adjust rates, sometimes in ways that make a more capable option cost more per task. Longer inputs, such as pasting entire contracts into a tool, raise costs further. Hidden expenses also accumulate outside the invoice itself, including data preparation, integration work, security reviews, and the internal time needed to test whether outputs are accurate. Because these costs sit in different departments' budgets, nobody sees the full picture. Set spending caps and alerts with each vendor, requiring teams to tag usage to named projects, and measuring cost per outcome rather than total spend. Finance forecasts in ranges.
Anthropic
Anthropic builds the Claude family of large language models, focused on safety, helpfulness, and honesty.
www.anthropic.com ↗Anthropic's Sonnet-class model for everyday coding and agentic work, a direct upgrade to Sonnet 5 with vision, tool use and a 1M-token context window.
Anthropic's most agentic Sonnet — frontier-level coding and tool use at Sonnet speed and price, with vision.
Anthropic's Opus model for long-running agentic coding and knowledge work, with always-on adaptive thinking, vision, and a 1M-token context window.
Anthropic's Fable model for the most demanding reasoning and long-horizon agentic work, with always-on adaptive thinking, vision, and a 1M-token context window.
Anthropic's Claude 5-family model for the most demanding reasoning, coding, and agentic work — with vision and a 128K output budget.
Anthropic's most capable model — frontier reasoning, coding, and agentic tool use with vision.