TLDR: Anthropic has launched Claude Haiku 4.5, a new lightweight AI model offering performance comparable to Claude Sonnet 4 at one-third the cost and more than twice the speed. Designed for real-time, low-latency tasks, it features an optional ‘extended thinking mode’ and is optimized for chat assistants, coding, and multi-agent workflows.
Anthropic has officially released Claude Haiku 4.5, its latest artificial intelligence model, positioning it as a game-changer for applications demanding both high performance and cost efficiency. Launched on October 15, 2025, the model is now broadly available, offering capabilities akin to the previously state-of-the-art Claude Sonnet 4, but at a significantly reduced operational cost and enhanced speed.
Claude Haiku 4.5 is engineered to deliver comparable performance to Claude Sonnet 4, yet it operates at one-third the cost and more than twice the speed. This makes it particularly attractive for real-time, low-latency applications such as chat assistants, customer service agents, and pair programming. The model is accessible through the Claude API, as well as major cloud platforms including AWS Bedrock and Google Cloud Vertex AI, with pricing set at an economical $1 per million input tokens and $5 per million output tokens.
One of the standout features of Claude Haiku 4.5 is its hybrid reasoning system, which includes an optional ‘extended thinking mode.’ While the model defaults to rapid responses, users can activate this mode to allow the AI additional time for deeper consideration before generating an answer. This functionality, a notable upgrade from its predecessor, Claude Haiku 3.5, also provides users with access to the model’s reasoning process, offering transparency into its decision-making. Anthropic describes Haiku 4.5 as a model with ‘a combination of speed and intelligence that make it particularly effective at coding tasks and computer use.’
In terms of performance benchmarks, Claude Haiku 4.5 achieved a 73.3% score on SWE-bench Verified, a prominent coding benchmark, and 41% on Terminal-Bench. These results place it competitively alongside models like GPT-5 and Gemini 2.5, though it is noted to lag behind Anthropic’s more advanced models, Sonnet 4.5 and Opus 4.1, in these specific metrics. Despite this, its efficiency and speed are expected to drive innovation in multi-agent AI workflows, where it can serve as a fast, cost-effective sub-agent or orchestrator within larger systems.
The model was trained on a proprietary dataset that integrates publicly available internet information up to February 2025, non-public third-party data, contributions from data-labeling services and paid contractors, opt-in user data from Claude users, and internally generated data. Anthropic emphasizes its rigorous data cleaning and filtering techniques, including deduplication and classification, applied during the training process.
Also Read:
- Google Labs’ Jules: Autonomous AI Coding Agent Redefines Software Development
- The Pivotal Shift: AI Development Now Centered on Inference and Real-World Application
Industry feedback highlights the model’s potential to redefine AI workload distribution. As one user enthusiastically stated, ‘I’ve never built apps so fast, and it does super well. I don’t even need Claude Sonnet anymore.’ Anthropic itself noted, ‘Speed is the new frontier for AI agents operating in feedback loops. Haiku 4.5 proves you can have both intelligence and rapid output.’ This strategic positioning aims to establish Claude Haiku 4.5 as a new baseline for businesses seeking scalable AI solutions without compromising on quality, enabling them to optimize AI spend by reserving premium models only for the most complex tasks.
