Claude Haiku 4.5

The fast, cheap Claude: near-frontier quality at the lowest current-generation price, built for high-volume routing, extraction, classification, and moderation.

Claude Haiku 4.5 is Anthropic's budget tier — the model you reach for when you need Claude-quality results at high volume and low latency, and don't need the depth of Sonnet or Opus. At $1 in / $5 out per million tokens it's five times cheaper than Sonnet and twenty-five times cheaper than Opus, while still delivering capable, near-frontier intelligence for the kinds of tasks that make up a surprising share of real workloads.

Those tasks are the sweet spot: classification, routing (deciding which bigger model or path a request should take), data extraction, summarization, and content moderation — high-throughput jobs where speed and cost-per-call matter more than maximum reasoning depth. It carries a 200K context windowThe maximum amount of text the model can "see" at once — prompt plus prior conversation plus any documents you give it. Measured in tokens (which are roughly three-quarters of a word each). A 128K context window is about 96,000 words of input — roughly a 400-page book. Larger context windows let the model work with bigger documents but cost more to run. (smaller than Opus/Sonnet's 1M) and the same image-input capability.

A fair caveat: while Haiku is the cheapest current-generation Claude, it isn't the cheapest model on the market — budget tiers from OpenAI, Google, and especially open or DeepSeek models can undercut it. You're paying a modest premium for Claude's quality and ecosystem.

• High-volume, latency-sensitive work: classification, routing, extraction, summarization, moderation.
• A cheap first-pass model in a multi-model pipeline (Haiku triages, bigger models handle the hard cases).
• Cost-sensitive production tasks that still want Claude-quality output.
• Simple coding and tool-use at scale.

• Complex reasoning or hard coding — use Claude Sonnet 4.6 or Claude Opus 4.8.
• Very long documents needing a 1M-tokenThe basic unit of text a model reads and writes. Tokens are roughly three-quarters of a word in English — so 100 tokens is about 75 words. Models don't see letters or words directly; they see tokens. Pricing is almost always quoted per million tokens, and context windows are measured in tokens rather than words. window — Haiku is 200K.
• The absolute lowest price — open models and DeepSeek undercut it.
• Self-hosting — Claude is closed.

Proprietary and API-only, on Anthropic's standard commercial terms: commercial rights to your outputs, none to the model, accessed via Anthropic's API or Bedrock/Vertex. Enterprise and zero-retention options exist. No open weightsThe numerical values inside a trained model that encode everything it has learned. A model is, functionally, a giant list of weights — tens of billions of numbers for a mid-sized model, hundreds of billions for a frontier model. "Open-weight" means those numbers are published. "Downloading the weights" means getting the actual file you'd need to run the model yourself. or self-hosting, as with all Claude models.

Against Claude Sonnet 4.6, Haiku is far cheaper and faster but less capable — the right tool for volume, not depth. Against OpenAI's GPT-5.4 Mini and Google's Gemini 3.5 Flash, Haiku competes on quality-per-dollar in the budget tier, though those rivals (and cheaper open models) can win on raw price. The usual pattern is to use Haiku as the cheap workhorse and escalate only the requests that need more.