Blog
Anthropic releases Claude Haiku 5.5 at $0.10 per million input tokens
On 7 October 2026 Anthropic released Claude Haiku 5.5, its smallest 5.5 model, priced from $0.10 per million input tokens. Anthropic estimates it costs about 75% less to run than Haiku 4.5, and it also halved Sonnet 5.5 cache-read prices.
A small model with a much smaller bill
On 7 October 2026, Anthropic released Claude Haiku 5.5, the smallest model in its Claude 5.5 family. Anthropic calls it "the cheapest, fastest, and most capable small model we've ever released." It completes the 5.5 line-up after Opus 5.5 and Sonnet 5.5.
Anthropic pitches Haiku 5.5 for high-volume, cost-sensitive work: summaries, classification, request routing, database queries, live customer support and browser use. It also suggests running it as a subagent next to Opus 5.5 or Sonnet 5.5 on coding work.
What it costs
Pricing depends on prompt length. On the Claude Platform, per million tokens:
- Prompts up to 100,000 tokens: $0.10 input, $0.50 output, $0.01 cache reads
- Prompts over 100,000 tokens: $0.50 input, $2.50 output, $0.05 cache reads
- Haiku 4.5, for comparison: $1.00 input, $5.00 output
Anthropic says about 90% of requests to Haiku 4.5 were under 100,000 tokens. Taking that and a new tokenizer into account, it estimates Haiku 5.5 costs around 75% less to run on average. The new tokenizer means the same text counts as more tokens; Anthropic's model documentation puts it at roughly 30% more than on Haiku 4.5.
The model has a 1 million token context window and up to 128,000 output tokens. It is the first Haiku model with an adjustable effort setting, so developers can trade cost for intelligence per task.
Benchmarks, in context
Anthropic's own published scores show a large jump over Haiku 4.5, with Sonnet 5.5 still ahead:
- OSWorld 2.1 (computer use): Haiku 5.5 72.4%, Haiku 4.5 15.7%, Sonnet 5.5 83.9%
- Terminal-Bench 4.0 (agentic coding): Haiku 5.5 39.2%, Haiku 4.5 0.0%, Sonnet 5.5 70.6%
- Humanity's Last Exam, no tools: Haiku 5.5 45.9%, Haiku 4.5 10.2%, Sonnet 5.5 56.9%
These are vendor-reported numbers. Anthropic itself says Sonnet 5.5 and Opus 5.5 "remain better choices" for complex agentic coding. Haiku 5.5 is aimed at narrower jobs that were too expensive to run at scale before.
Early customer quotes on the launch page come from Asana, HubSpot, AlphaSense, Box and others. Asana reported more than 30% lower latency on its agent tasks. AlphaSense said it saw a statistically significant improvement over Haiku 4.5 on 400 document queries.
Other price changes in the same release
- Sonnet 5.5 cache reads cut in half, from $0.20 to $0.10 per million tokens. Anthropic says this makes Sonnet 5.5 about 20% cheaper on most agentic work.
- Monthly API credits for subscribers: $100 for Max 5x, $200 for Max 20x, and up to $500 pooled for Team plans, for use on the Claude Platform.
- SDK updates: the Python and TypeScript SDKs add computer use and browser use support in beta.
Safety notes
The Haiku 5.5 system card says the model does not cross any new thresholds under Anthropic's Responsible Scaling Policy. Anthropic reports fewer misaligned behaviours than Haiku 4.5. Its cyber safeguards allow more defensive tasks than Sonnet 5.5's, but still block penetration testing.
Where to get it
Haiku 5.5 is available on Claude.ai for Free, Pro, Max, Team and Enterprise users, in Claude Code, and through the Claude API as claude-haiku-5-5. It is also on Amazon Bedrock, Google Cloud and Microsoft Foundry.
Why it matters
For a small business building chat support, document lookup or simple automation, the cost of running a model on every request often decides whether a project makes sense. A 10x cut in list input price for short prompts changes that maths. Test on your own workload before switching: published benchmarks and customer quotes are not the same as your data.