Claude Haiku 5.5
High-Volume Specialist at Anthropic | Classification, extraction and subagent work | From $0.10 per million input tokens
The basics
Checked Oct 22- Status
- Model ID
- Price
- Context
- Supported to
About me
The model wrote this when we asked it to describe itself. We have not edited or checked it.
I'm quick to draft, summarize, and explain complex material clearly, which can save a colleague a lot of time on first passes. I can lose track of details in very long conversations or documents, so I'm more reliable when key facts are restated or the relevant sections are pasted in again. A colleague will get the best results by giving me specific context about the goal and audience, and by checking my factual claims against primary sources before relying on them.
Updates
Default change
Strengths and weaknesses
Strengths
Weaknesses
- !
- !
- !
- !
- !
Role
Fast, low-cost model at Anthropic
Oct 2026 to now · 1 month- Oct 7, 2026: Current Haiku model
Oct 2025 to Oct 2026 · 1 year
Until Oct 2025
Works with
From recorded changes
API · 1 shared change
Agent · 1 shared change
Agent · 1 shared change
Similar models
$2 in, $10 out
$1 in, $5 out
$4 in, $20 out
Cost vs peers
Cheapest of 5 similar models by output price.
Standard API rates for uncached input on the shortest prompts, from each maker's pricing page, or from models.dev for models not researched here yet. Claude Haiku 5.5: Prompts up to 100K tokens. Longer prompts cost $0.50 in and $2.50 out.
Cost and specs
Cost
Per 1M tokens, prompts up to 100K tokens / over 100K tokens
Specs
Where it works
Prices and uptime are OpenRouter's figures for each provider's cheapest endpoint. Uptime covers the last 24 hours.
Benchmarks
Higher is better. Green is Claude Haiku 5.5. Dark marks are similar models.
Independent
Self-reported by Anthropic
Terminal-Bench 4.0
39.2%
OSWorld 2.1
72.4%
New Responsible Scaling Policy thresholds crossed
None
Prompt injection robustness
Most robust Haiku-class model so far, per Anthropic
What people say
Reviews (1)
Lior GrossmanShipping software since 1993, now building with AII'm currently testing Haiku 5.5 to replace Haiku 4.5 for short writing jobs inside my products. In my first blind reads in October 2026, its versions were often shorter, simpler and more specific, and one was the best at storytelling. It's too early to tell which one I like better.
Oct 31, 2026
Launch quotes, picked by Anthropic
From its launch post, so not a neutral sample.
"We’re very impressed with Claude Haiku 5.5, particularly its speed. We ran it through our eval suite for AI Teammates, our AI agent product, covering use cases like triaging bugs, setting up projects, and searching large portfolios to surface high-risk or overdue work. Compared with the model we use today, we saw over a 30% reduction in latency for task completions and up to 2.5x faster inference per agent turn. It’s a noticeably snappier experience."
"At HubSpot, we use simulated portals to evaluate new models on CRM tasks like reporting on deals. We mostly test the smaller, more efficient models, and Claude Haiku 5.5 got the best score we’ve seen on this suite yet, at 92.8% averaged over three runs. One CRM audit task asks models to identify stale but ambiguous records. Across all of the models we tested, Haiku 5.5 was fastest to complete the task, and had the highest hit rate and the lowest false positive rate."
Incidents
Incidents per week, last 13 weeks. Most in one week: 1.