Robots Atlas>ROBOTS ATLAS
Claude Haiku 4.5

Claude Haiku 4.5

Claude Haiku 4.5 (claude-haiku-4-5-20251001) · Family: Claude
Anthropic’s fastest model (Claude family, 2025): matches Sonnet 4 on coding and computer use at ~1/3 the cost; SWE-bench 73.3%; 200K context, extended thinking.
✓ Active✓ Public accessLLMMultimodalReasoning modelTool-using model📁 Claude
Context window
200K
tokens
Max output
64,000
tokens
Release date
15 October 2025
Access:APIHostedDeployment:☁ Cloud

Overview

Claude Haiku 4.5 is Anthropic’s fastest model in the Claude family, released on 15 October 2025. It is a lightweight variant with near-frontier intelligence, offered at a lower price and high speed.

The model matches Claude Sonnet 4 on coding, computer use and agent tasks at about one-third the cost and more than twice the speed; it also runs up to 4–5× faster than Claude Sonnet 4.5. On SWE-bench Verified it scores 73.3% (with a 128K thinking budget).

It is multimodal — it accepts text and images and produces text. It supports a 200,000-token context window and up to 64,000 output tokens, as well as extended thinking. Its reliable knowledge cutoff is February 2025 (training data through July 2025).

API pricing is USD 1 per 1M input tokens and USD 5 per 1M output tokens. The model is available via the Claude API, Claude Code, Amazon Bedrock and Google Vertex AI. It was released under the AI Safety Level 2 (ASL-2) standard. Typical uses include fast real-time applications, customer service, multi-agent systems and coding sub-agents.

Classification
LLMMultimodalReasoning modelTool-using model
Family: Claude
Access & deployment
APIHosted
Cloud
Weights: Closed
Key parameters
📏 Context: 200K
Tools
📥 Input: text, image, documents

Technical specification

Context window
200K
tokens
Max output tokens
64,000
tokens per response
Knowledge cutoff
1 Feb 2025
Knowledge boundary
License
Proprietary
Features:Tool use
Modalities
⬇ Input
textimagedocuments
⬆ Output
textcodestructured_data

Capabilities and applications

Native model capabilities
Reasoning
The model's ability to reason logically and solve complex problems.
Category: reasoning
Multi-step reasoning
Carrying out multi-step chains of reasoning across long, complex tasks.
Category: reasoning
Long context
Support for large context windows — tens to hundreds of thousands (or millions) of input tokens. Enables analysis of entire codebases, long documents, and many parallel conversations without losing earlier information. GPT-5.1 supports 400,000 tokens.
Category: language
Coding
Generating, analysing and modifying code in many programming languages. Covers writing functions, debugging, refactoring, code review, and creating tests. Measured by benchmarks such as HumanEval and SWE-bench.
Category: coding
Agentic coding
Multi-hour, multi-step programming tasks performed autonomously by the model: cloning a repository, running tests, iterating on fixes, integrating with CLI tools. Characteristic of Codex variants (GPT-5.1-Codex-Mini, Codex-Max).
Category: coding
Tool use
The model's ability to call external functions, APIs and tools during a conversation: calculator, search engine, code editor, database. The model decides when and how to use a tool and interprets its result.
Category: planning
Structured output
Producing data in structured formats such as JSON.
Category: structured_generation
Image understanding
Analysing and interpreting the content of images.
Category: vision
Chart understanding
Reading and interpreting charts, tables and diagrams.
Category: vision
OCR
Recognising text within images and documents.
Category: vision
Multilingual
Competence in many natural languages (from a few to over a hundred): understanding, generation, translation, and code-switching within a single conversation. Frontier models support a wide range of languages with comparable quality.
Category: language
Computer use
The model's ability to operate a computer interface by interpreting screenshots and generating actions such as clicks, typing, and navigating applications.
Category: planning
Extended thinking mode
A reasoning-model variant with a larger inference budget: more thinking cycles, higher answer precision at the cost of response time. Choice between 'standard' and 'extended' thinking is left to the user (e.g. the selector in GPT-5.2 Pro).
Category: reasoning

Benchmark results

1 benchmark
SWE-bench
accuracy · SWE-bench Verified, 128K thinking budget
73.3%
📄 Anthropic (Claude Haiku 4.5, X 2025)

Pricing

Technical architecture

Deployment and security

🔒 Security / Enterprise
✓ Verified enterprise information
Updated: 24 Jul 2026↗ Security documentation