Claude Haiku 4.5 (latest)

Claude Haiku 4.5 sits at the lightweight end of Anthropic's lineup, positioned by the company as a fast lane for responsive chat, lightweight agent workflows, and routine office tasks. It belongs to the claude-haiku family and was made available on the same day it was announced, suggesting a streamlined rollout aimed at developers who need quick iteration rather than frontier reasoning depth. The model accepts text, image, and PDF inputs and produces text outputs, while weights remain closed and the model is served through Anthropic's own API as well as a broad set of nineteen distribution routes listed in third-party catalogs, which makes it easy to slot into existing pipelines and gateway setups.

In practice, the model is shaped around practical deployments that benefit from low latency and modest cost. It supports tool calling, adjustable temperature, and reasoning, which together make it suitable for agents that orchestrate external services, summarize documents, or handle back-and-forth dialog. the cataloged API limit context window paired with the cataloged API limit output ceiling supports moderately long documents and multi-turn interactions, while pricing sits at the affordable end of Anthropic's catalog, helping teams route high-volume, lower-stakes traffic to Haiku while reserving larger models for harder problems.

Anthropicclaude-haiku-4-5claude-haiku

Open provider page Provider docs

Quick Info

Provider: Anthropic
Model key: claude-haiku-4-5
Release date: Oct 15, 2025
Last updated: Oct 15, 2025
Knowledge cutoff: 2025-02-28
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $1.00
Output token cost: $5.00

Limits

Output tokens: 64,000 tokens
Context window: 200,000 tokens

Transparent token rates

Compare Claude Haiku 4.5 (latest) pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about Claude Haiku 4.5 (latest)

Anthropic

Official sourceAnnouncement

Claude Haiku 4.5, our latest small model, is available today to all users.

Anthropic

CoverageBenchmark

Benchgraph, a third-party benchmark aggregation site, lists Claude Haiku 4.5 (latest) with a SWE-bench Verified score of 73.3 dated April 2026. The score is drawn from ModelSpec cards and reflects daily research on the 500-task human-validated subset of Princeton's SWE-bench, which measures end-to-end resolution of rea The benchgraph listing confirms that Haiku 4.5 remained the active Haiku variant as of April 2026 and continues to be tracked against major coding benchmarks. The page aggregates 112 models with scores and is part of a free, open-source project from the team behind Dark Product Factories, with plans to expand into per-

Anthropic

CoverageBenchmark

benchgraph.dev explicitly lists "Claude Haiku 4.5 (latest)" with provider "anthropic" and reports a SWE-bench Verified score of 73.3 dated 2026-04, placing Haiku 4.5 in the coding/agentic benchmark context behind Claude Mythos Preview (93.9), Claude Opus 4.6 (80.8), Gemini 3.1 Pro Preview (80.6), and o3-pro (73.2), and Beyond the single score row, benchgraph.dev provides infrastructure-level context that is useful for Haiku 4.5 evaluation: it tracks 112 models today, groups 28 coding benchmarks (plus 6 agentic) under a per-benchmark page model refreshed by daily research, and ties benchmark evidence into its ModelSpec cards so downst