Claude Haiku 4.5

Claude Haiku 4.5 is designed as a lightweight powerhouse that brings frontier-level capability to real-time and high-volume applications. It introduces extended thinking to the Haiku family, enabling controllable reasoning depth with summarized or interleaved thought output. The model matches Claude Sonnet 4's performance on coding, computer use, and agent tasks, delivering near-frontier intelligence at a fraction of the cost and latency of larger Claude models. Its architecture supports full tool-assisted workflows including coding, bash, web search, and computer-use tools, making it suitable for complex multi-step tasks where responsiveness matters.

The model achieves a 73.3% score on SWE-bench Verified, ranking it among the world's best coding models while maintaining exceptional responsiveness for sub-agents, parallelized execution, and scaled deployment. Built to serve developers across multiple platforms including Claude Code, Cursor, Amazon Bedrock, Google Vertex AI, and Microsoft Foundry, Haiku 4.5 is engineered for high-volume, cost-sensitive production environments. Its extended thinking capabilities allow developers to dial reasoning depth up or down depending on task complexity, while prompt caching and batch processing features support efficient resource utilization at scale.

Perplexity Agentanthropic/claude-haiku-4-5claude-haiku

Open provider page Provider docs

Quick Info

Provider: Perplexity Agent
Model key: anthropic/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 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

Perplexity Agent

CoverageBenchmark

Magica's model catalog page for Claude Haiku 4.5 restates core specifications: 200K input context, 64K output limit, $1.00 per million input tokens, $5.00 per million output tokens, with an official release date of October 15, 2025 and cache read/write rates of $0.10 and $1.25 per million tokens respectively. It notes The page adds third-party benchmark indices sourced from Artificial Analysis: Intelligence Index 16.9, Coding Index 43.9, Agentic Index 8, and a Best Design Arena score of 1,159 Elo (with a 41-rank placement in Asciiart at 49.3% win rate). Pricing add-ons of $0.01 per web search are also listed, providing a consolidate

FrogBot

CoverageBenchmark

InfoQ's November 2025 release coverage headline-captures the core pitch for Claude Haiku 4.5: faster performance at approximately one-third the cost of Anthropic's larger frontier models. The piece frames the launch as a continuation of Anthropic's rapid late-2025 shipping cadence, in which Sonnet 4.5 and Haiku 4.5 arr For developers and platform teams, InfoQ positions Haiku 4.5 as a model that meaningfully shifts the build-vs-buy and model-routing calculus: a single Claude variant that is cheap enough for high-volume serving, fast enough for interactive pair-programming, and capable enough (per Anthropic's published benchmark claims

Perplexity Agent

CoverageBenchmark

The BenchLM aggregator profile for Claude Haiku 4.5, with data as of September 10, 2026, records the model as released October 15, 2025, proprietary and non-reasoning, with a 200,000-token context window. API pricing is listed at $1 per million input tokens and $5 per million output tokens, with cached input at $0.10 a Haiku 4.5 receives a composite capability score of 52.8 out of 100 on BenchLM, ranking 103rd out of 232 models. Its strongest eligible category is Knowledge at the 37th percentile (a raw rank value of 116), described as particularly effective for knowledge-intensive workloads like research, analysis, and factual Q&A. A