AI Tools
Claude Haiku vs Sonnet: which Claude model should you pay for?
Claude Haiku 5.5 vs Sonnet 5.5 is a 20x price gap: Haiku costs $0.10 per million input tokens and $0.50 per million output tokens, while Sonnet 5.5 costs $2.00 and $10.00. Anthropic launched Haiku 5.5 on October 7, 2026 and cut Sonnet's cache-read price the same day.
What matters
- Claude Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens, per Anthropic's pricing page.
- Above 100,000 tokens of prompt the Haiku 5.5 price steps up fivefold, to $0.50 input and $2.50 output per million tokens.
- Anthropic prices Haiku 5.5 90% below Haiku 4.5 under 100k tokens and 50% below above it, though it states the average saving is around 75% once the new tokenizer is counted.
- On Anthropic's own Terminal-Bench 4.0 numbers, Sonnet 5.5 scores 70.6% to Haiku 5.5's 39.2%, so the cheap model is not a drop-in replacement for agentic coding.
- Sonnet 5.5 cache reads fell from $0.20 to $0.10 per million tokens on launch day, cutting its cost on most agentic tasks by roughly 20%.
What is Claude Haiku 5.5?
Claude Haiku 5.5 is Anthropic's small, low-cost model, released on October 7, 2026 and identified in the API as claude-haiku-5-5. Anthropic describes it as the cheapest, fastest, and most capable small model it has released.
It is built for high-volume, repetitive work: summaries, compaction, database queries, classification, and extraction. It is also the first Haiku-class model with an adjustable effort setting, so you can trade cost against accuracy on a per-request basis instead of picking one setting for everything.
Claude Haiku 5.5 vs Sonnet 5.5 pricing (verified October 2026)
Anthropic prices both models per million tokens, and the gap is the whole story. On prompts up to 100,000 tokens, Haiku 5.5 charges $0.10 for input and $0.50 for output. Sonnet 5.5 charges $2.00 and $10.00, which is twenty times more on both input and output.
Haiku 5.5 has one wrinkle that catches people out. The low rate only applies to prompts up to 100,000 tokens; above that the price rises fivefold to $0.50 input and $2.50 output. Anthropic says prompts under 100,000 tokens covered around 90% of requests to the previous Haiku model, so most chat and classification workloads never cross the line, but a long-document or codebase-wide pipeline will.
Cache reads are where the two models nearly meet: $0.01 per million tokens on Haiku 5.5 against $0.10 on Sonnet 5.5. Sonnet's cache-read rate was halved from $0.20 to $0.10 on the same day as the Haiku launch.
How much cheaper is Haiku 5.5 than Haiku 4.5?
Against the model it replaces, the cut is sharp. Haiku 4.5 costs $1.00 per million input tokens and $5.00 per million output tokens. Haiku 5.5's $0.10 and $0.50 are 90% lower for prompts up to 100,000 tokens, and 50% lower above that.
Anthropic states the blended saving is around 75% less to run, not 90%, because Haiku 5.5 uses an updated tokenizer similar to Sonnet 5.5's and so consumes slightly more tokens for the same task. If you are modelling spend, treat the per-token price as the headline and expect a modest token-count bump on top.
The honest catch: when Sonnet 5.5 is still worth 20x
Cheaper is not better at everything. Anthropic states plainly that Sonnet 5.5 and Opus 5.5 remain the better choice for complex agentic coding. In the vendor's published benchmark table, Haiku 5.5 scores 39.2% on Terminal-Bench 4.0 against Sonnet 5.5's 70.6%, and on GDPval-AA v2.1 knowledge work Haiku 5.5 records 1620 versus Sonnet 5.5's 1840 and Haiku 4.5's 735.
The pattern to hold onto is scope. Haiku 5.5 fits narrowly-scoped, high-volume steps such as compaction, summarisation, classification, and subagent work inside a larger agent. It is a poor fit for long, multi-step coding sessions, where one weak step compounds and a failed run costs more than the tokens you saved.
Availability and the new Max and Team API credits
Haiku 5.5 is available now on Amazon Web Services, Google Cloud, and Microsoft Azure, and on the Claude Platform under the model ID claude-haiku-5-5. Anthropic also updated its Python and TypeScript SDKs to add computer-use and browser-use support in beta, tasks the model's speed and price suit.
Alongside the launch, Anthropic is adding monthly Claude Platform credits for subscribers: $100 per month for Max 5x, $200 for Max 20x, and up to $500 pooled for Team. The credits can be spent on any Claude model and are meant for building agents and apps that call the API.
Claude Haiku 5.5 vs Sonnet 5.5: which should you pay for?
- Pick Haiku 5.5 if your work is high-volume and repetitive, or if you are bolting a cheap subagent onto a larger pipeline. The 20x saving is real and the quality drop rarely shows on narrow tasks.
- Pick Sonnet 5.5 if you run agentic coding, long tool-use chains, or anything where a single bad step ruins the run. Paying 20x for a task Sonnet finishes first time beats retrying it with Haiku.
- Re-test before staying on Haiku 4.5. At $1.00 and $5.00 per million tokens it now costs ten times Haiku 5.5 for prompts under 100k, so price alone is a reason to move once your eval suite passes.
At a glance
| Price per 1M tokens | Claude Haiku 5.5 | Claude Haiku 4.5 | Claude Sonnet 5.5 |
|---|---|---|---|
| Input (prompt up to 100k tokens) | $0.10 | $1.00 | $2.00 |
| Input (prompt over 100k tokens) | $0.50 | $1.00 | $2.00 |
| Output (prompt up to 100k tokens) | $0.50 | $5.00 | $10.00 |
| Output (prompt over 100k tokens) | $2.50 | $5.00 | $10.00 |
| Cache read | $0.01 up to 100k / $0.05 over | $0.10 | $0.10 |
| Best fit | High-volume narrow tasks and subagents | Previous Haiku generation | Complex agentic coding |
FAQ
How much does Claude Haiku 5.5 cost?
Haiku 5.5 costs $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens. Above 100,000 tokens the rate rises to $0.50 input and $2.50 output per million, per Anthropic's pricing page.
Is Claude Haiku 5.5 better than Claude Sonnet 5.5?
Only on price and speed. Haiku 5.5 is 20x cheaper per token, but Anthropic's own benchmark table puts Sonnet 5.5 ahead on complex agentic coding, 70.6% versus 39.2% on Terminal-Bench 4.0. Use Haiku for narrow, high-volume tasks and Sonnet for agentic coding.
Where can I use Claude Haiku 5.5?
It is available now on Amazon Web Services, Google Cloud, and Microsoft Azure, and on the Claude Platform under the model ID claude-haiku-5-5. Anthropic also added computer-use and browser-use support to its Python and TypeScript SDKs in beta.
Related reading
Claude tool profile, Claude API pricing, Cheapest LLM API, Claude Haiku pricing: what does Haiku 5.5 cost per million tokens?