Frontier flagship models
Claude Haiku 5.5
Made by Anthropic
Anthropic's small model for high-volume work, with adjustable effort and paid API access — released October 7, 2026.
What it is
Claude Haiku 5.5 launched October 7, 2026 for high-volume, cost-sensitive applications such as summaries, classification and live customer support, and for coding subagents alongside Sonnet 5.5 and Opus 5.5.
Anthropic calls it its fastest model and reports around 75% lower average running cost than Haiku 4.5. Haiku 5.5 is the first Haiku with adjustable effort, trading cost against capability. These are Anthropic's claims; no independent benchmark score is assigned here.
For prompts up to 100K tokens, API input costs $0.10 per million tokens and output $0.50; above 100K input, these rise to $0.50 and $2.50. Cache reads cost $0.01 / $0.05 and cache writes $0.125 / $0.625 per million tokens for the same two prompt-length bands.
Use the official announcement and provider documentation for availability. Paid API access does not establish a free Claude.ai allowance, and this library has no Haiku endpoint enabled.
Paid API access is not a free consumer plan. Higher effort and prompts over 100K tokens can increase costs.
Published price
- API — input (up to 100K / over 100K)$0.10 / $0.50 per 1M tokens
- API — output (up to 100K / over 100K input)$0.50 / $2.50 per 1M tokens
- API — cache reads (up to 100K / over 100K)$0.01 / $0.05 per 1M tokens
- API — cache writes (up to 100K / over 100K)$0.125 / $0.625 per 1M tokens
Prices depend on prompt length. API access is paid; this announcement does not establish free Claude.ai access to Haiku 5.5. Provider prices may differ.
Source: Anthropic's October 7 launch announcement · Last checked 8 October 2026
Benchmarks
No numerical benchmark results are recorded here for this model. Maker-published evaluations are not independent results.
Where to read today's numbers
Scores move, so these are the boards that keep them current.
- Anthropic's Haiku 5.5 evaluationsMaker's own figures
Launch evaluations published by Anthropic, the model's maker; not independent benchmark results.
Published capabilities
Released October 7 for high-volume, cost-sensitive work. Anthropic describes Haiku 5.5 as its fastest model and says it costs around 75% less to run than Haiku 4.5 on average. It is the first Haiku with adjustable effort; these are maker-reported claims, not independent results.
Prompting it from this site
Can't be run from here
This library has no Haiku endpoint enabled. Use the Claude Platform or a supported cloud provider; API use is paid.
Use it for
- Summarization and classification
- Live customer support
- Coding subagents alongside Sonnet and Opus
Running it yourself
Hosted only — nothing to install
Claude Haiku 5.5 is hosted only; local code can call its paid API, but there are no weights to install.
What you need first
- A Claude Platform account with API billing, or supported cloud-provider access
- For direct API use, an API key stored privately in your environment
Step by step
1.Follow the Claude Platform quick start or your cloud provider's setup guide
2.For the Claude Platform, select model claude-haiku-5-5
3.Choose an effort setting for the task and check the prompt-length pricing band
Anthropic lists availability on AWS, Google Cloud and Microsoft Azure as well as the Claude Platform. Provider prices can differ; this library has no Haiku endpoint enabled.