Claude Haiku 5.5 is Anthropic’s October 7, 2026 small model and the last piece of the Claude 5.5 family. It is built for high-volume, cost-sensitive work — summaries, compactions, database queries, classification — and, as Anthropic’s fastest model to date, for latency-bound jobs like live customer support and browser use. It also slots in as a cheap subagent beneath Claude Opus 5.5 and Claude Sonnet 5.5.
The price is the headline: $0.10 / $0.50 per million tokens for prompts up to 100K, about 75% less than Haiku 4.5 on average — with a capability jump that moves it well past what a “small” Claude used to do.
Benchmarks
Vendor-reported (Anthropic, October 7, 2026). Sonnet 5.5 is shown “for reference”; GPT-6 Luna figures are Anthropic-run.
| Benchmark | What it measures | Haiku 5.5 | Haiku 4.5 | GPT-6 Luna | Sonnet 5.5 |
|---|---|---|---|---|---|
| GDPval-AA v2.1 | Real-world knowledge work (Elo) | 1620 | 735 | 1437 | 1840 |
| AA-Briefcase v1.1 | Professional deliverables (Elo) | 1578 | 614 | 1336 | 1824 |
| OSWorld 2.1 (offline subset) | Computer use | 72.4% | 15.7% | 48.9% | 83.9% |
| Humanity’s Last Exam | Reasoning (no tools / tools) | 45.9% / 57.4% | 10.2% / 18.7% | — | 56.9% / 64.5% |
| Terminal-Bench 4.0 | Agentic terminal coding | 39.2% | 0.0% | 16.4% | 70.6% |
| FrontierCode 1.1 (Main) | Mergeable agentic coding | 46.4% | — | 42.4% | 52.1% (xhigh) |
| Chartography | Chart recognition (no tools) | 46.4% | 6.4% | 29.1% | 61.6% |
Haiku 5.5 beats GPT-6 Luna on every row Anthropic published, but it sits clearly below Sonnet 5.5 — and Anthropic is explicit that Sonnet 5.5 and Opus 5.5 remain the better choice for complex agentic coding. Haiku 5.5 is “best suited to more narrowly scoped tasks.”
Customer results line up with that positioning: Asana saw over a 30% latency reduction and up to 2.5x faster inference per agent turn; HubSpot’s CRM suite gave Haiku 5.5 its best small-model score at 92.8%; Box measured an 11-point gain over Haiku 4.5 at about half the latency; and Cognition uses it as the “sidekick” in Devin Fusion, holding a 66.2 FrontierCode score while cutting cost and latency.
Pricing
| Price per 1M tokens | Haiku 5.5 (prompts up to 100K) | Haiku 5.5 (over 100K) | Haiku 4.5 | Sonnet 5.5 |
|---|---|---|---|---|
| Input | $0.10 | $0.50 | $1.00 | $2.00 |
| Output | $0.50 | $2.50 | $5.00 | $10.00 |
| Cache reads | $0.01 | $0.05 | $0.10 | $0.10 |
| Cache writes | $0.125 | $0.625 | $1.25 | $2.50 |
Anthropic says prompts up to 100K tokens made up around 90% of requests to the previous Haiku, which is where Haiku 5.5 is “especially good value.” Alongside the launch it also halved Sonnet 5.5’s cache-read price to $0.10 per million and introduced a monthly API credit for Claude Max and Team subscribers.
Specs
| Spec | Value |
|---|---|
| API model ID | claude-haiku-5-5 |
| Context window | 1M tokens |
| Max output | 128K tokens |
| Thinking | Adaptive; adjustable effort (default Medium) — a first for Haiku |
| Reliable knowledge cutoff | June 2026 |
| Modalities | Text and image in, text out |
| Platforms | Claude Platform, Amazon Web Services, Google Cloud, Microsoft Azure |
Safety and safeguards
- Alignment. Major improvements over Haiku 4.5 on almost every alignment evaluation, with far fewer misaligned behaviors and a lower willingness to cooperate with misuse.
- Cybersecurity. Safeguards are more restrictive than Haiku 4.5’s but somewhat less restrictive than on Anthropic’s other recent models; they permit a wider range of defensive tasks than Sonnet 5.5’s while still blocking penetration testing and other attacker-leaning techniques.
- Biology. The same safeguards as Sonnet 5, Sonnet 5.5 and Opus 5. Organizations can apply to the Life Sciences Verification Program and Cyber Verification Program for broader access.
How Haiku 5.5 compares
- vs Claude Haiku 4.5: roughly a tenth of the input price, a much larger context window and effort control, plus a step change on every benchmark. A straightforward migration for most workloads.
- vs Claude Sonnet 5.5: a fifth of the input price but about half the agentic-coding score — use Haiku for scoped, high-volume work and Sonnet for harder tasks.
- vs GPT-6 Luna: Luna is priced at the same $0.10 / $0.50, and Anthropic’s own comparison favors Haiku 5.5 on every row it ran.
Limitations
- Context pricing cliff. Above 100K tokens the price is 5x higher, so long-document pipelines lose most of the savings.
- Narrow scope. The model is meant for focused tasks; do not substitute it for Sonnet or Opus on open-ended agentic coding.
- Self-reported numbers. The GPT-6 Luna scores are from Anthropic’s harness, not OpenAI’s.
For the wider picture, see the 2026 LLM guide.
Best for
- Classification, extraction, routing, summaries and database queries at high volume
- Fast subagent beneath Opus 5.5 or Sonnet 5.5 on coding work
- Live customer support and browser-use agents where latency matters
- Document Q&A at scale (AlphaSense reports 0.84 vs 0.76 for Haiku 4.5 on 400 queries)
Pros & cons
Strengths
- Costs around 75% less than Haiku 4.5 on average — $0.10 / $0.50 per 1M tokens for prompts up to 100K
- First Haiku-class model with an adjustable effort setting, so you can trade cost for intelligence per request
- Big capability jump over Haiku 4.5 — Terminal-Bench 4.0 39.2% (from 0.0%), OSWorld 2.1 offline 72.4% (from 15.7%), GDPval-AA 1620 (from 735)
- Anthropic's fastest model to date; customers report 30%+ lower latency and up to 2.5x faster inference per agent turn (Asana)
Limitations
- Not for complex agentic coding — Terminal-Bench 4.0 39.2% vs Sonnet 5.5's 70.6%; Anthropic points to Sonnet or Opus for those jobs
- Prompts over 100K tokens are priced 5x higher ($0.50 / $2.50), eroding the cost advantage on long context
- Cybersecurity safeguards block penetration testing and other attacker-leaning techniques, though they allow more defensive work than Sonnet 5.5's
- Vendor-reported benchmarks only; GPT-6 Luna comparisons come from Anthropic's own harness
Frequently asked questions
When was Claude Haiku 5.5 released?
October 7, 2026, completing the Claude 5.5 family after Opus 5.5 (September 22) and Sonnet 5.5 (September 28).
How much does Claude Haiku 5.5 cost?
$0.10 per 1M input tokens and $0.50 per 1M output tokens for prompts up to 100K tokens ($0.50 / $2.50 above 100K), versus $1 / $5 for Haiku 4.5. Cache reads are $0.01 per 1M and cache writes $0.125. Anthropic puts the average saving at around 75%.
What is new compared with Haiku 4.5?
Haiku 5.5 is the first Haiku with an adjustable effort setting (default Medium), a 1M-token context window with up to 128K output, and far stronger agentic scores — Terminal-Bench 4.0 39.2% vs 0.0%, OSWorld 2.1 offline 72.4% vs 15.7%, Humanity's Last Exam 45.9% vs 10.2% without tools.
Should I use Haiku 5.5 or Sonnet 5.5?
Use Haiku 5.5 for narrowly scoped, high-volume, latency-sensitive work such as classification, extraction, routing and subagents. Use Sonnet 5.5 or Opus 5.5 for complex agentic coding — on Terminal-Bench 4.0 Sonnet 5.5 scores 70.6% against Haiku's 39.2%.
What is the Claude API model ID for Haiku 5.5?
`claude-haiku-5-5`. It is available on the Claude Platform, Amazon Web Services, Google Cloud and Microsoft Azure.