Claude Haiku 5.5 is now in every model chooser on your CRHQ server. Anthropic calls it the cheapest, fastest and most capable small model it has released. It replaces Haiku 4.5: anything that used Haiku 4.5 now uses Haiku 5.5 automatically, with nothing for you to change.
A big step up from Haiku 4.5
These are Anthropic's reported results. The chart compares Haiku 5.5 with the Haiku it replaces and with GPT-6 Luna, OpenAI's small model. Sonnet 5.5 is shown for reference.
| Benchmark | What it measures | Haiku 5.5 | Haiku 4.5 | GPT-6 Luna | Sonnet 5.5 |
|---|---|---|---|---|---|
| GDPval-AA v2.1 | Knowledge work | 1620 | 735 | 1437 | 1840 |
| AA-Briefcase v1.1 | Knowledge work | 1578 | 614 | 1336 | 1824 |
| OSWorld 2.1 | Computer use | 72.4% | 15.7% | 48.9% | 83.9% |
| Humanity's Last Exam (no tools) | Multidisciplinary reasoning | 45.9% | 10.2% | n/a | 56.9% |
| Humanity's Last Exam (with tools) | Multidisciplinary reasoning | 57.4% | 18.7% | n/a | 64.5% |
| Terminal-Bench 4.0 | Agentic coding | 39.2% | 0.0% | 16.4% | 70.6% |
| FrontierCode 1.1 (Main) | Agentic coding | 46.4% | n/a | 42.4% | 52.1% |
| Chartography | Visual reasoning | 46.4% | 6.4% | 29.1% | 61.6% |
The jump over Haiku 4.5 is large everywhere. On knowledge work the GDPval-AA rating more than doubles, from 735 to 1620, and on computer use the OSWorld score goes from 15.7% to 72.4%. Haiku 5.5 is also the first Haiku with an adjustable effort setting, so it can be tuned for cost or for intelligence.
Anthropic is also clear about where the bigger models still win: Sonnet 5.5 and Opus 5.5 remain the better choice for complex agentic coding. Haiku 5.5 is best at narrowly scoped work. Our suggestion is to use it for quick, high-volume jobs such as summaries, lookups, classification and helper agents, and to keep Sonnet 5.5 or Opus 5.5 for the heavy lifting.
Pricing
| Per 1M tokens | Haiku 5.5, prompts up to 100k | Haiku 5.5, prompts 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 |
Haiku 5.5 is priced by prompt length. Prompts up to 100,000 tokens get the lower price, and Anthropic says those made up around 90% of requests to the previous Haiku. Longer prompts still cost half of what Haiku 4.5 did.
One thing to know for long sessions: the 100,000 tokens are counted for each request, and the whole prompt counts, including context read from or written to the cache. In long agent sessions most requests are over 100,000 tokens, so the saving there is about half. Short, focused jobs get the full saving, about a tenth of the old price.
Haiku 5.5 has a 1 million token context window and can write up to 128,000 tokens in one reply. With this launch Anthropic also halved the price of Sonnet 5.5 cache reads, from $0.20 to $0.10 per million tokens.
What changes on your CRHQ server
- Two Haiku options in every model chooser. Haiku 5.5 is pinned, so your choice stays on exactly this version. Haiku (latest) always follows the newest Haiku, which is now 5.5.
- Haiku 4.5 is retired in CRHQ. Haiku 5.5 is far better at a lower price, so CRHQ no longer offers the old one.
- The switch is automatic. Everything that was set to Haiku, including agents, sessions and scheduled jobs, now runs Haiku 5.5. You do not need to change anything. Older chats keep the model and cost they were recorded with.
- Effort control now works for Haiku. Turn it down for cheap, fast answers or up for harder work. The default is medium.
Pricing is tracked per turn in each session's Costs tab, and Claude subscription rotation works exactly as before.
Also in this release
Agents can run longer commands. A command an agent runs can now take up to 10 minutes by default, and up to 18 minutes when the agent asks for more. A command that passes its limit is stopped and the agent is told. Work that needs longer than that is best set up as a scheduled job, which runs on its own and reports back when it is done.
Availability
| Fact | Value |
|---|---|
| Availability | Every CRHQ server |
| Live since | 2026-10-08 |
| Provider | Claude |
| Pricing | From $0.10 in / $0.50 out per 1M tokens |
| Client action required | None |