Claude Haiku 5.5: Anthropic cuts prices for high-volume AI work
Anthropic has released Claude Haiku 5.5, a small model for high-volume, cost-sensitive work that it says costs about 75% less than Haiku 4.5 on average. A sourced look at pricing, benchmarks and availability for enterprises. As of October 8, 2026.
CogLake Editorial · 8 October 2026 · 4 min
This article was automatically generated with the assistance of AI.
On October 7, 2026, Anthropic released Claude Haiku 5.5, its new small model for high-volume, cost-sensitive tasks — and cut entry pricing sharply: prompts up to 100,000 tokens now cost $0.10 per million input tokens and $0.50 per million output tokens. Anthropic says Haiku 5.5 runs about 75% cheaper on average than Haiku 4.5. Here is what is confirmed, what remains vendor-reported, and what the release means for companies running many small AI tasks.
What is confirmed (as of October 8, 2026)
- Announcement: Published October 7, 2026 on Anthropic's product page (source below).
- Availability: API identifier
claude-haiku-5-5. Anthropic states the model is available now "on all platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure." - Tiered pricing: $0.10/$0.50 per 1M input/output tokens for prompts up to 100k; $0.50/$2.50 above that.
- Effort setting: the first Haiku-class model with an adjustable effort setting (optimize for cost or intelligence).
- Related changes: Claude Sonnet 5.5 cache reads halved to $0.10 per 1M tokens (about 20% cheaper on most agentic work, per Anthropic); new monthly API credits for Claude Max and Team subscribers ($100 / $200 / up to $500 pooled), rolling out this week; Python and TypeScript SDKs gain beta support for computer use and browser use.
Prices at a glance
| Price per 1M tokens (USD) | Haiku 5.5, prompt ≤ 100k | Haiku 5.5, prompt > 100k | Haiku 4.5 |
|---|---|---|---|
| Cache reads | $0.01 | $0.05 | $0.10 |
| Cache writes | $0.125 | $0.625 | $1.25 |
| Input | $0.10 | $0.50 | $1.00 |
| Output | $0.50 | $2.50 | $5.00 |
Source: Anthropic announcement of October 7, 2026. Anthropic calculates a 90% lower input price than Haiku 4.5 for requests up to 100k tokens and 50% lower above that; it also notes that Haiku 5.5's updated tokenizer uses slightly more tokens per task.
Benchmarks: vendor-reported, not independent tests
| Benchmark | Haiku 5.5 | Haiku 4.5 | Sonnet 5.5 |
|---|---|---|---|
| GDPval-AA v2.1 (knowledge work) | 1620 | 735 | 1840 |
| OSWorld 2.1, offline subset (computer use) | 72.4% | 15.7% | 83.9% |
| Humanity's Last Exam, no tools (reasoning) | 45.9% | 10.2% | 56.9% |
| Terminal-Bench 4.0 (agentic coding) | 39.2% | 0.0% | 70.6% |
All figures come from Anthropic's announcement and the system card linked there; we have not run our own measurements. Anthropic continues to position Sonnet 5.5 and Opus 5.5 as the better choices for complex agentic coding, while Haiku 5.5 targets narrowly scoped, high-frequency tasks such as summarisation, compaction and subagent work. Customer statements (for example Asana reporting over 30% lower task-completion latency compared with a previous model) are also vendor-reported.
What this means for enterprise workloads
- Cheaper bulk work. Classification, extraction, summaries, triage and routing are dominated by per-token cost — exactly the tier Haiku 5.5 is built for.
- Subagent pattern. Anthropic explicitly suggests pairing Haiku 5.5 with Opus 5.5 or Sonnet 5.5 as a subagent for coding work: the small model handles the frequent steps, the larger model the hard ones.
- Watch the tier boundary. Pricing changes at 100k prompt tokens. Check where your workloads sit and factor in the tokenizer change.
- Safety and verification. Haiku 5.5's cybersecurity safeguards are more restrictive than Haiku 4.5's but permit more defensive tasks than Sonnet 5.5's, per Anthropic. Teams with regulatory requirements should read the system card and the Life Sciences / Cyber Verification Programs.
- Availability in your stack. An announcement is not the same as usability in your tools. CogLake connects models vendor-neutrally, and Anthropic is part of its provider portfolio; whether and when Haiku 5.5 is available in your workspace depends on the provider connection and its approval in your deployment — see the current model list at coglake.com/ai-models.
If you are evaluating models for document-heavy work, our article on EmbeddingGemma 2 covers the retrieval side of the same question.
Sources
- Anthropic: "Introducing Claude Haiku 5.5", October 7, 2026 — https://www.anthropic.com/claude-haiku-5-5 (retrieved October 8, 2026).
- Prices and benchmark figures are Anthropic statements as of October 7, 2026 and can change.