NEWEST SONNET1M CONTEXTCACHE READS HALVED OCT 7THINKING ON BY DEFAULTTEXT + VISION
Claude Sonnet 5.5 API Pricing
Claude Sonnet 5.5 replaced Claude Sonnet 5 as Anthropic's current Sonnet on September 28, 2026 at Sonnet 5's rate card, and on October 7 Anthropic halved its cache read. It now costs $2/M input, $10/M output and $0.10/M cache-hit input - 5% of input instead of 10% - which Anthropic says makes it about 20% cheaper on most agentic work. Cache writes, batch and the tokenizer match Sonnet 5. What can move the bill the other way is thinking: it is on by default at high effort, reasoning tokens bill as output, and disabled is no longer accepted. Checked by hand against platform.claude.com.
Input - per 1M tokens
$2.00/M
Unchanged from Sonnet 5 = Sonnet 5
Output - per 1M tokens
$10.00/M
Includes thinking tokens = Sonnet 5
Cached input - 95% off
$0.10/M
Cache write $2.50 / 5 min -95%
Effective - agentic blend
$1.29/M
92/8 split - 82% cache
§ 01 / TERMINAL
Run the numbers.
Live calculator pre-loaded with Claude Sonnet 5.5 rates. Adaptive thinking is on by default, so put your expected reasoning tokens in the reasoning field - they bill at the output rate. The lowest thinking setting, between_tools, switches off up-front thinking, and effort (default high) sets how much thinking there is.
$ /mo
Workload split
Prompt cache hit rate
Tokens you can process
—
Words equivalent (English)
—
Effective rate
—
§ 02 / SCENARIOS
Real-world presets.
SUPPORT AGENT
Ticket triage with tools
$0.021/ticket
CODING AGENT
Repo-wide bug fix
$0.184/task
LONG DOC ANALYSIS
Contract risk review
$0.208/doc
RESEARCH OPS
Analyst brief draft
$0.072/brief
Estimate · anthropic-bpe-estimate · ≈2.6 chars/token Auto-counts as you type
This is a chars-per-token approximation, not a real tokenizer. Actual tokens vary by language, code density, and tool-call overhead — counts are typically ±10–20% off for English prose, more for code or non-Latin scripts. For exact billing, use the vendor's official tokenizer.
Characters 916
Words 153
Tokens (estimated) 352 tokens
Cost as input · uncached $0.0007 USD
Cost as output · uncached $0.00352 USD
Cost as cached input $0.00004 USD
| Model | Input /M | Output /M | Effective blended | Context | Best for |
|---|---|---|---|---|---|
| Claude Sonnet 5.5 Current | $2.00 cache $0.10 | $10.00 | $1.29 agentic 92/8 | 1M | Speed-and-intelligence default |
| Claude Sonnet 5 | $2.00 cache $0.20 | $10.00 | $1.36 +6% | 1M | The model Sonnet 5.5 replaces, 2x the cache read |
| Claude Opus 5.5 | $4.00 cache $0.20 | $20.00 | $2.58 +2.0× | 1M | Anthropic's recommended default, twice the rate |
| Claude Haiku 5.5 | $0.10 cache $0.01 | $0.50 | $0.0682 −19× | 1M | Support and classification volume |
| Claude Fable 5.1 | $10.00 cache $0.25 | $50.00 | $6.26 +4.9× | 1M | Anthropic's top capability tier |
| GPT-6.1 Sol | $2.00 cache $0.10 | $10.00 | $1.29 same | 1.05M | OpenAI model at the same list price |
| Grok 4.7 | $2.00 cache $0.50 | $6.00 | $1.19 −8% | 500K | Cheaper output for coding agents |
| Gemini 3.8 Flash | $0.75 cache $0.075 | $3.75 | $0.481 −2.7× · promo | 1M | Mainstream multimodal work |
Frequently asked.
Practical Claude Sonnet 5.5 pricing questions - what the rate card says, what the October 7 cache-read cut changed, and where the bill can differ from Sonnet 5.
Q · 01 What is Claude Sonnet 5.5 priced at? +
Anthropic lists
claude-sonnet-5-5 at $2/M input, $0.10/M cache-hit input and $10/M output. Cache writes are $2.50/M for the 5-minute window and $4/M for the 1-hour window, and batch is $1/M and $5/M.Q · 02 Is Sonnet 5.5 more expensive than Sonnet 5? +
No - since October 7, 2026 it is slightly cheaper. It launched on Sonnet 5's exact rate card, and Anthropic's migration page still says the prices are the same, but on October 7 Anthropic halved Sonnet 5.5's cache read to
$0.10/M; Sonnet 5 keeps $0.20/M. Every other line is identical and so is the tokenizer, so the same text produces the same token count. Under our standard blend (92% input, 82% cache hits) that is $1.29/M here against $1.36/M for Sonnet 5. The bill can still rise for a different reason: thinking is on by default, and reasoning tokens bill at the output rate.Q · 03 Can I turn thinking off to cut the bill? +
Not completely. On Sonnet 5.5
thinking: {"type": "disabled"} returns a 400 error. The lowest setting is between_tools, which switches off up-front thinking; it is accepted at low, medium and high effort and rejected at xhigh and max. Manual thinking budgets also return a 400. Default effort is high, and Anthropic says effort levels were recalibrated, so a setting carried over from Sonnet 5 does not produce the same amount of thinking.Q · 04 How is cached input billed? +
Cache hits cost
$0.10/M, 5% of the input price - Anthropic cut it from 10% ($0.20/M) on October 7, 2026, and says the change makes Sonnet 5.5 about 20% cheaper on most agentic work. Writes are unchanged at $2.50/M for a 5-minute cache and $4/M for a 1-hour cache, and they replace the base input price rather than adding to it, so a 5-minute cache pays for itself after one hit. The minimum cacheable prompt is 512 tokens.Q · 05 Is there a Batch API discount? +
Yes, 50% on input and output:
$1/M and $5/M. On the Message Batches API, Sonnet 5.5 can also return up to 300k output tokens with the output-300k-2026-03-24 beta header, against 128k on the synchronous API.Q · 06 Does the full 1M context cost extra? +
No. Anthropic includes the full 1M-token window at standard pricing on Claude 4.6 and later - a 900k-token request bills at the same per-token rate as a 9k-token one - and prompt caching and batch discounts apply across the whole window. The window is both the default and the maximum.
Q · 07 Does US-only inference change the price? +
Yes. Setting
inference_geo: "us" applies a 1.1x multiplier to every token category - input, output, cache writes and cache reads - on Claude 4.6 and later models. The prices on this page use the default global routing. Bedrock and Google Cloud set their own regional prices.Q · 08 How does it compare with Claude Opus 5.5 on price? +
Opus 5.5 lists at
$4.00/M input and $20.00/M output, twice this model. Both now read cache at 5% of their own input price, so Opus 5.5's cache read, $0.20/M, is also twice this model's $0.10/M. Anthropic's own guidance is to start with Opus 5.5 for most workloads; it describes Sonnet 5.5 as the best combination of speed and intelligence, with faster comparative latency.Q · 09 What breaks when migrating from Sonnet 5? +
Anthropic lists five breaking changes, none of them in the price.
thinking: disabled is replaced by between_tools; forced tool use returns an error; thinking blocks are tied to the model and the conversation; on the Claude API and Google Cloud the older computer_20251124 tool is not accepted; and the advisor tool rejects Opus 4.8, Opus 4.7 and Sonnet 5 as advisors. Separately, text between tool calls now comes back inside thinking blocks, which can silence a progress stream.Q · 10 What model ID should developers use? +
claude-sonnet-5-5 on the Claude API, Google Cloud, Microsoft Foundry and Claude Platform on AWS, and anthropic.claude-sonnet-5-5 on Amazon Bedrock. Anthropic commits to retirement not sooner than September 28, 2027 on the platforms it operates.