Claude Sonnet 5.5 pricing: $2 in, $10 out

According to APITariff, Claude Sonnet 5.5 pricing on 2026-10-15 is $2 per 1M input tokens and $10 per 1M output tokens, with cache hits at $0.10 per 1M and the Batch API at $1 and $5, all as listed on Anthropic’s pricing page. Take an assumed workload of 1,000,000 calls that each send 1,000 input tokens and return 300 output tokens, not a measured average. The bill reads $5,000 at list price and $2,500 on Batch. The reseller kie.ai lists the model at $0.80 and $4.00, 60% lower, which bills $2,000, though that saving shrinks on calls that reuse a cached prompt. Sonnet 5.5 has no long-prompt surcharge: a 900,000-token prompt and a 9,000-token prompt pay the same per-token rate.

Last updated 2026-10-15 · 6 models compared · sources: Anthropic’s pricing page and the OpenRouter and kie.ai rows our fetch read, 2026-10-15.

Claude Sonnet 5.5 pricing, field by field

Every Sonnet 5.5 price on Anthropic’s pricing page, read 2026-10-15, is below. The Batch API takes 50% off both token prices. The US-only row is 1.1x standard, billed when a request sets US-only inference.

Field ($ per 1M tokens)	Claude Sonnet 5.5
Input	2.00
Output	10.00
5-minute cache write	2.50
1-hour cache write	4.00
Cache hit	0.10
Batch input / output	1.00 / 5.00
US-only inference, input / output	2.20 / 11.00

Anthropic’s page lists fast mode for “Claude Opus 5.5, Claude Opus 5, and Claude Opus 4.8” only, so Sonnet 5.5 has no fast-mode rate. Bedrock, Vertex and Azure rates are not covered here and were not fetched.

How much does Claude Sonnet 5.5 cost at 1M calls?

Claude Sonnet 5.5 costs $5,000 for 1,000,000 calls that each send 1,000 input and 300 output tokens. The figures in this table are computed from the rates above and that assumed workload: $2,500 on Batch, $7,500 on Claude Sonnet 4.6, $10,000 on Claude Opus 5.5 and $250 on Claude Haiku 5.5. This table ranks price, not quality, and carries no score.

Model	Input / output $ per 1M	Bill at 1M calls
Claude Haiku 5.5	0.10 / 0.50	$250
Claude Sonnet 5.5, Batch	1.00 / 5.00	$2,500
Claude Sonnet 5.5	2.00 / 10.00	$5,000
Claude Sonnet 5	2.00 / 10.00	$5,000
GPT-6.1 Sol	2.00 / 10.00	$5,000
Claude Sonnet 4.6	3.00 / 15.00	$7,500
Claude Opus 5.5	4.00 / 20.00	$10,000

Source: Anthropic pricing page for the Claude rows, OpenRouter API for GPT-6.1 Sol, all read 2026-10-15. Haiku 5.5 is the rate for prompts up to 100,000 tokens, and it bills more above that. For other call volumes, use the token calculator.

Is Claude Sonnet 5.5 cheaper than Sonnet 4.6?

On list price, yes: $2 and $10 against $3 and $15, 33.3% lower on both. The tokenizer narrows the gap. On Anthropic’s pricing page, Claude 4.7 and later models count “approximately 30% more tokens for the same text” than Sonnet 4.6 and earlier, which keep the older tokenizer. If your text turns into 30% more tokens, Sonnet 5.5’s $2 input is $2.60 per 1M old-tokenizer-equivalent tokens and its $10 output is $13.00, 13.3% below Sonnet 4.6 on both. The 30% is Anthropic’s approximation, and the real figure depends on your content.

Sonnet 5.5 and Claude Sonnet 5 list the same $2 and $10. They differ on cache hits: $0.10 on Sonnet 5.5 and $0.20 on Sonnet 5, per the same pricing page, so a cached prompt costs half as much on Sonnet 5.5.

How much does prompt caching save on Sonnet 5.5?

Prompt caching cuts a Sonnet 5.5 call that reads 50,000 cached tokens, adds 10,000 fresh input tokens and returns 1,000 output tokens from $0.13 to $0.035, 73.1% lower, when the cache is warm and no write cost applies. Without caching, the same 60,000 input tokens cost $0.12 and the whole call $0.13. Ten thousand such calls cost $350.

Anthropic halved Sonnet 5.5’s cache-read price with the Haiku 5.5 launch on 2026-10-07, from $0.20 to $0.10. Its launch page says “We’re halving the price of Claude Sonnet 5.5’s cache reads” and that Sonnet 5.5 now runs around 20% cheaper on most agentic work. On the call above the halving takes the bill from $0.040 to $0.035, 12.5% lower, so this call sits under Anthropic’s 20% figure.

Did the Sonnet 5.5 price change?

Input and output prices did not change in our fetches. We hold 16 price rows for OpenRouter’s anthropic/claude-sonnet-5.5 slug from 2026-09-29, our first fetch of it, to 2026-10-15, and every row reads $2 and $10. The kie.ai row has read $0.80 and $4.00 in all 15 price rows since 2026-10-01. The cache-read price did change, per Anthropic’s launch page above. Our fetch holds input and output prices only, so it cannot show that change.

Is Claude Sonnet 5.5 cheaper on kie.ai?

Yes at list price, and not always on cached prompts. kie.ai charges $0.80 per 1M input tokens and $4.00 per 1M output tokens for Claude Sonnet 5.5, 60% below Anthropic’s $2.00 and $10.00. That is $2,000 against $5,000 per 1,000,000 calls of the assumed workload. Its Sonnet 5.5 page, read 2026-10-15, prints the same two rates and says both are 60% below official pricing.

Route	Input / output $ per 1M	Bill at 1M calls
kie.ai	0.80 / 4.00	$2,000
Anthropic Batch API	1.00 / 5.00	$2,500
Anthropic list	2.00 / 10.00	$5,000

kie.ai is 20% under Batch, and Batch is asynchronous. Treat them as list prices. kie.ai bills in credits, and we have not looked at its rate limits, uptime or terms.

Where kie.ai is not the cheapest: its Sonnet 5.5 page prints no cache rate, so we price every input token at $0.80. If kie.ai caches, the gap narrows. The 60,000-token call above then bills $0.052 on kie.ai against $0.035 direct, which puts direct 32.7% lower. Direct beats kie.ai once more than 68.4% of a call’s input tokens are cache hits, at 60,000 input and 1,000 output tokens. The standard workload has no cache hits, so kie.ai wins there. The same page says high-tier top-ups lower the effective price by about 10%. We did not use that figure. A different model is cheaper still: GPT-6.1 Sol on kie.ai lists $0.60 and $3.00, which bills $1,500 per 1,000,000 calls of the assumed workload.

Affiliate link. If you sign up through it we may earn a commission. It changed nothing in the numbers above, which were fetched before the link existed. Get Claude Sonnet 5.5 on kie.ai.

Does Sonnet 5.5 charge more for long prompts?

No. Anthropic’s pricing page, read 2026-10-15, says “Claude 4.6 and later models (except Claude Haiku 5.5) and Claude Mythos Preview include the full 1M token context window at standard pricing”, and that caching and batch discounts apply across the whole window. Haiku 5.5 is the exception: it bills more once a prompt passes 100,000 tokens. xAI works the other way. Once a Grok 4.7 prompt reaches 200,000 tokens, every token in the request bills at double rate, as the Grok API pricing guide prints from xAI’s page of 2026-10-15. A 250,000-token prompt on Sonnet 5.5 costs $0.50 for input at $2 per 1M.

The Anthropic API pricing guide covers every current Claude model, the Claude vs OpenAI comparison puts them next to OpenAI’s tiers, and the Claude Opus 5.5 post has the next tier up. If you hold a Claude subscription, Claude Max vs API cost sets it against these rates. How these figures are fetched is on the methodology page.

Cite as: APITariff, apitariff.com/guides/claude-sonnet-5-5-pricing/, fetched 2026-10-15.

Drafted with AI from APITariff’s own price data, checked and edited by a person. The chart is rendered from the tables above, not generated.

Change log

• 2026-10-15First edition. Anthropic's rates read on its pricing page 2026-10-15, OpenRouter and kie.ai rows fetched the same day.