Introducing Claude Haiku 5.5: the cheapest, fastest, and most capable small model we’ve ever released. On average, it costs around 75% less to run than Claude Haiku 4.5.
1,609
3,275
45,252
6,677,280
Haiku 5.5 is a significant step up over Haiku 4.5 across coding, computer use, and knowledge work.
156
348
6,414
887,058
Haiku 5.5 is designed for high-volume, cost-sensitive tasks. It reliably handles repetitive work like summaries and classification, and pairs well with Claude Opus 5.5 and Sonnet 5.5 as a subagent on coding work. It’s also fast enough for live customer support and browser use.
32
54
3,482
282,575
Haiku 5.5 is our first Haiku model with an adjustable effort setting, so you can decide whether to optimize for cost or intelligence on each task.

Oct 7, 2026 · 6:01 PM UTC

35
66
2,381
273,135
Haiku 5.5 is especially good value on tasks under 100,000 tokens, which make up around 90% of requests to our previous Haiku model.
39
84
2,836
371,971
Haiku 5.5 shows major improvements across almost all of our alignment evaluations relative to Haiku 4.5, with far fewer instances of misaligned behavior.
19
27
1,646
208,636
Haiku 5.5 is available now on all platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure. Read more: anthropic.com/claude-haiku-5…

Introducing Claude Haiku 5.5

Claude Haiku 5.5 is our fastest, most capable small model. Built for high-volume work like summarization, subagents, and browser use.

anthropic.com
17
36
1,592
315,285
One more thing: we’re halving the price of cache reads on Claude Sonnet 5.5, to $0.10 per million tokens. That makes Sonnet 5.5 around 20% cheaper to run on most long-running work.
218
590
11,868
1,563,681
Sort replies: Relevant Recent Liked
Replying to @claudeai
Anthropic is mogging the entire industry at this point
6
1,021
Replying to @claudeai
Oh f****** now you're going to start screwing up the user interface to give people more control and then blame us when we're getting bad results? Where's your data on people asking for this. I suppose you want me to do my own research on healthcare next
342
Replying to @claudeai
💨
bidyut
@bidyut_io
Oct 7
Opus after compression 💨
2
614
Replying to @claudeai
@grok can you explain each graph what it says about?
1
47
Replying to @claudeai
It can’t do much, but what it can do, it’s better at?
133
Replying to @claudeai
A dial for how hard it thinks, picked per task. I love that. Low for sorting a pile of receipts, high for the one email you really can't get wrong. You stay in charge of what you spend. Nice work on this one. Best, Pete
2
1,099
Replying to @claudeai
Finally charts humans can read.
205
Replying to @claudeai
Useful knob. It also changes how to compare 4.5 and 5.5 on your own tasks: effort is one more variable, so fix it and repeat the run. A dev.to test today ran one Claude Code skill on Haiku 5.5 three times, and the no-skill baseline scored 0.30, 0.85 and 0.61.
42
Replying to @claudeai
Will it auto-route to the most effecient and cost-effective model? Having people (the masses) choose is where it would fail because they have no idea what to choose or how much a task would cost.
1
278
Replying to @claudeai
Haiku max for Opus medium intelligence
2
220
Replying to @claudeai
Ahh we never knew we needed this great
479
Replying to @claudeai
Adjustable effort per task makes more sense than a flat model tier. Most requests don't need full intelligence, they need a fast cheap answer, and only occasionally need the model to actually think harder. Nice to get that control per call instead of per model.
886
Replying to @claudeai
So, it looks like Haiku 5.5 high is comparable on many tasks to Sonnet 5.5 medium for a lower cost. Sounds good to me.
1
618
Replying to @claudeai
Claude family dinner probably looks like: Opus: “I handle the hard stuff.” Sonnet: “I handle most things.” Haiku: “I cost 75% less.” everyone: “…”
595
Replying to @claudeai
we need auto-model selection
1
1
399
Replying to @claudeai
That adjustable effort setting is a really interesting new feature.
AGTP
@AGTPinsights
Oct 7
Anthropic just launched Claude Haiku 5.5, its cheapest and fastest small model. Here's what you need to know. On average, Haiku 5.5 costs around 75% less to run than Haiku 4.5. Anthropic calls it its cheapest, fastest and most capable small model yet. It's a big step up over Haiku 4.5 in coding, computer use and knowledge work. Anthropic also reports far fewer misaligned behaviors in its alignment evaluations. It's the first Haiku with an adjustable effort setting, so you can tune each task for cost or intelligence. The context window is 1M tokens, up from 200K. Output goes up to 128K tokens, up from 64K. The price has a cutoff at 100K tokens. Prompts under 100K make up about 90% of requests to the previous Haiku. Under that line, Haiku 5.5 is around 10x cheaper than Haiku 4.5. Cursor says the same for shorter requests. Commentators note the under-100K price matches OpenAI's GPT-6 Luna. It's built for high-volume, cost-sensitive work like summaries, classification, compactions and database queries. Anthropic says it's fast enough for live customer support and browser use. It also works as a subagent under Opus 5.5 or Sonnet 5.5. Key numbers: - Under 100K tokens: $0.10 input, $0.50 output per million, cache reads $0.01 - Over 100K tokens: $0.50 input, $2.50 output per million, cache reads $0.05 - Sonnet 5.5 cache reads halved to $0.10 per million, about 20% cheaper on most long-running work - New monthly API credits: Max 5x $100, Max 20x $200, Team up to $500 pooled Haiku 5.5 is live today, Oct 8, in the Claude Platform, Claude Code, AWS, Google Cloud, Microsoft Azure and Cursor. It's the third Claude 5.5 model in a month.
297