Claude Haiku 5.5 Is Here: Anthropic's Cheapest, Fastest Model Yet
Anthropic announced Claude Haiku 5.5 on October 7, 2026: around 75% cheaper to run than Haiku 4.5, its fastest model to date, and the first Haiku with an adjustable effort setting. What changed, what the benchmarks say, and why the unit economics of AI features just shifted.
Amir Ali Liaqat · Founder & CEO, DesignsToDeploy
On October 7, 2026, Anthropic announced Claude Haiku 5.5, which it describes as the cheapest, fastest, and most capable small model it has ever released. The headline is as much about price as capability: Anthropic says Haiku 5.5 now costs around 75% less to run than Haiku 4.5, while posting benchmark scores the previous generation could not touch.
Haiku 5.5 is designed for high-volume, cost-sensitive work: summaries, compactions, database queries, classification, the quick repetitive tasks that pile up inside real products. Anthropic's intended pattern is to pair it with Opus 5.5 and Sonnet 5.5 as a subagent on coding work: the small model handles the volume, the big models handle the hard reasoning. And because it is Anthropic's fastest model to date, it is also aimed at speed-sensitive jobs like live customer support and browser use.
The Price Story
The pricing is two-tiered. For prompts up to 100,000 tokens, input costs $0.10 per million tokens and output $0.50, compared with $1.00 and $5.00 on Haiku 4.5. Over 100K tokens the rates step up to $0.50 input and $2.50 output. Anthropic notes that around 90% of requests to the previous Haiku model stayed under 100K tokens, so most workloads land in the cheaper tier.
Alongside the launch, Anthropic halved the price of Claude Sonnet 5.5's cache reads to $0.10 per million tokens, which it says makes Sonnet 5.5 around 20% cheaper on most agentic work. It also introduced a new monthly API credit for Max subscribers ($100 to $200) and Team subscribers (up to $500, pooled), intended to support building new agents and applications on the Claude Platform.
The Benchmarks
The figures below are Anthropic's own, published with the announcement. On Terminal-Bench 4.0 for agentic coding, Haiku 5.5 scored 39.2% where Haiku 4.5 scored 0.0%. On OSWorld 2.1 for computer use, 72.4% against 15.7%. On GDPval-AA for knowledge work, 1620 against 735. Anthropic is careful to note that Sonnet 5.5 and Opus 5.5 remain the better choice for complex agentic coding, but for narrowly scoped tasks that used to be cost-prohibitive, the small model now carries real weight.
Tune It for Cost or Intelligence
Haiku 5.5 is the first Haiku-class model with an adjustable effort setting plus adaptive thinking, which defaults to medium. Like the larger models, you can now decide per task whether to optimize for cost or for intelligence. For product teams running agents at scale, that dial matters: it turns the model from a fixed price into a controllable budget line.
Where to Get It
Claude Haiku 5.5 is available now on all platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure. On the Claude Platform, developers can get started with the model ID claude-haiku-5-5. Anthropic is also updating its Python and TypeScript SDKs to add support for computer use and browser use in beta, and Haiku 5.5 is positioned as especially well suited to those tasks given its speed, capability, and price.
Why This Matters for Product Teams
When a capable small model costs this little per call, AI features that were budget line-items become affordable defaults. Support triage, browser agents, coding subagents, document processing, chatbots: the unit economics just shifted. The winners will not be the teams with the biggest model, but the teams that split work intelligently between cheap, fast models and heavy ones.
At DesignsToDeploy, we build AI features into real products: support assistants, automation, and agent workflows on modern stacks. Building AI into your product? Let's build it together: www.designstodeploy.dev
1 / 6Keep reading
Splentra Is Live: Split Expenses Smartly with Friends
Splentra, our expense-splitting app, is now live on Google Play. Create groups, split bills equally or with custom shares, and let the debt-simplification engine settle balances with the fewest payments.
Read article
Introducing Splentra: A Smarter Way to Split Expenses
Meet Splentra, our new expense-splitting app for friends, roommates, and colleagues — create groups, track shared expenses, and settle debts without the awkward math.
Read article
Claude Dashboards and Motion: Anthropic Turns Plain Language into Live Charts and Editable Animations
On October 8, 2026, Anthropic launched Claude Dashboards and Claude Motion in beta, and took Docs, Slides, and Design out of beta on every plan, including Free. How plain-language dashboards work, what Motion can animate, and why the standalone Claude Design site closes on December 14.
Read article