Claude Haiku 5.5: Around 75% Cheaper Than Haiku 4.5

Introduction

Anthropic launched Claude Haiku 5.5 on October 7, 2026 (model ID claude-haiku-5-5). Anthropic describes it as "the cheapest, fastest, and most capable small model we've ever released," aimed at high-volume, cost-sensitive work such as summaries, compactions, database queries and classification requests. It is also pitched as a subagent for Opus 5.5 and Sonnet 5.5 on coding work, and as a fit for speed-sensitive jobs like live customer support and browser use.

The headline change is price. Anthropic says Haiku 5.5 costs around 75% less to run than Haiku 4.5 on average. This article stays on Haiku 5.5; for the larger siblings, see our earlier coverage of Claude Sonnet 5.5 and Claude Opus 5.5.

Feature Overview

1. Much lower per-token pricing. Anthropic's pricing table lists two tiers by prompt size (up to 100,000 tokens, and over 100,000 tokens).

Anthropic's footnote explains the "around 75%" figure: Haiku 5.5 is priced 90% lower than Haiku 4.5 for requests up to 100,000 tokens and 50% lower above that. On Haiku 4.5, 90% of requests fell into the shorter category. The calculation also accounts for an updated tokenizer, similar to those of Sonnet 5.5 and Opus 5.5, which makes Haiku 5.5 use slightly more tokens per task.

2. Speed. Anthropic calls Haiku 5.5 its fastest model to date at each model's standard speed, and notes that it runs less quickly than the Opus models in Fast Mode.

3. An adjustable effort setting. Anthropic says Haiku 5.5 is its first Haiku-class model with an adjustable effort setting, so users can trade cost against intelligence. Its charts for OSWorld, GDPval-AA and Humanity's Last Exam plot accuracy against cost at Low, Med, High, Xhigh and Max.

4. Benchmarks. The table below is reproduced from Anthropic's post; columns are Haiku 5.5, Haiku 4.5, GPT-6 Luna and Sonnet 5.5, and a dash means no score was given.

The FrontierCode row carries an "Xhigh" annotation in Anthropic's table. Anthropic points readers to the Haiku 5.5 System Card for how evaluations were run.

5. Safety and safeguards. Anthropic says Haiku 5.5 shows major improvements across almost all of its alignment evaluations relative to Haiku 4.5, with far fewer instances of misaligned behavior and a lower willingness to cooperate with misuse. Its cybersecurity safeguards are more restrictive than Haiku 4.5's, somewhat less restrictive than those on other recent models, and permit a wider range of defensive tasks than the safeguards for Sonnet 5.5 while still blocking penetration testing. Biology safeguards match those for Sonnet 5, Sonnet 5.5 and Opus 5. Organizations can apply to the Life Sciences Verification Program and Cyber Verification Program.

Usability Analysis

Anthropic says Haiku 5.5 is available now on all platforms, including Amazon Web Services, Google Cloud and Microsoft Azure, and on the Claude Platform as claude-haiku-5-5. A migration guide is linked from the announcement. Anthropic is also updating its Python and TypeScript SDKs to add computer use and browser use in beta, and calls Haiku 5.5 well suited to those tasks.

Customer quotes in the post are early-testing reports, not independent tests. Asana reported over a 30% reduction in latency for task completions and up to 2.5x faster inference per agent turn compared with the model it uses today. HubSpot reported 92.8% averaged over three runs on its simulated CRM portal suite. AlphaSense measured 0.84 versus 0.76 for Haiku 4.5 over 400 queries on its "Ask in Document" feature, which it says handles about 8M calls a week in production. Box reported scoring 11 points higher than Haiku 4.5 at about half the latency. Cognition said that with Haiku 5.5 as the sidekick in Devin Fusion and Opus 5.5 as lead, Fusion holds a FrontierCode score of 66.2 while cutting cost and latency.

Anthropic is explicit about scope. Sonnet 5.5 and Opus 5.5 remain better choices for complex agentic coding such as Terminal-Bench 4.0, while Haiku 5.5 suits narrowly scoped tasks like compaction, summarization and subagent work that may previously have been cost-prohibitive.

Pros and Cons

The strengths are the steep price cut, a tiered cache price that falls to $0.01 per million tokens for shorter prompts, an effort dial, and wide platform availability. The limits are a large gap to Sonnet 5.5 on agentic coding (39.2% versus 70.6% on Terminal-Bench 4.0), a higher price tier for prompts over 100,000 tokens, and benchmark and customer figures that come from Anthropic and its partners.

Context From the Same Announcement

The same post also halves Sonnet 5.5's cache-read price (from $0.20 to $0.10 per million tokens), which Anthropic says cuts Sonnet 5.5's cost on most agentic tasks by around 20%. It also announces a monthly API credit for Max and Team subscribers: $100 for Max 5x, $200 for Max 20x, and up to $500 pooled for Team, usable on any model.

Outlook

The release reflects a pattern of pairing a larger lead model with cheaper subagents, a workflow Anthropic and Cognition both describe. If the quoted pricing holds in production, workloads such as classification, summarization and compaction that were marginal at Haiku 4.5 prices become easier to justify. Teams should verify results on their own prompts, since the tokenizer change means actual savings depend on token counts for each workload.

Conclusion

Claude Haiku 5.5 is a cost and speed release for high-volume, well-scoped work, backed by a new effort setting and broad platform availability. It is not positioned as a replacement for Sonnet 5.5 or Opus 5.5 on hard coding tasks. It is best suited to developers running subagents, support bots and bulk text pipelines. Rating: 4 out of 5.

Editor's Verdict

Claude Haiku 5.5: Around 75% Cheaper Than Haiku 4.5 earns a solid recommendation within the Claude space.

The strongest case for paying attention: Anthropic prices input at $0.10 per million tokens for prompts up to 100,000 tokens, versus $1.00 for Haiku 4.5. That alone raises the bar for what readers should expect in this space. Reinforcing that, adjustable effort setting lets teams tune cost against intelligence — practical value rather than just headline appeal. The broader signal worth registering is straightforward: price, not peak capability, is the main change: Anthropic keeps Sonnet 5.5 and Opus 5.5 as the choice for complex agentic coding and positions Haiku 5.5 for narrowly scoped work. On the other side of the ledger, one constraint is real rather than a marketing footnote: the gap to Sonnet 5.5 on agentic coding is wide, at 39.2% versus 70.6% on Terminal-Bench 4.0, and Anthropic itself recommends larger models for complex coding. It should factor into any serious decision. Layered on top of that, prompts over 100,000 tokens fall into a pricier tier ($0.50 input, $2.50 output), which narrows the saving to about 50% — which narrows the set of teams for whom this is an obvious yes.

For developers running high-volume subagents, support bots and bulk summarization or classification pipelines on Claude, this is a serious evaluation candidate, not just a curiosity to bookmark. For everyone else, the safer posture is to monitor coverage and revisit once the use cases that matter to your team are demonstrated in the wild.