Anthropic
Claude Opus 4.7 pricing
Anthropic's April 2026 flagship. Same $5/$25 rate card as Opus 4.6, with a retrained tokenizer and five-level thinking budget.
Input
$5.00/ 1M tok
Output
$25.00/ 1M tok
- Context window
- 1M
- Max output
- 128K
- Cached input
- -
- Verified
- 2026-04-17
Claude Opus 4.7 is Anthropic's April 2026 release and takes over as the top-end model in the Claude family. Input is $5 per 1M tokens, output is $25 per 1M tokens - identical to Opus 4.6 - but the underlying tokenizer is new. For the same block of text Opus 4.7 emits 1.0 to 1.35× more tokens than Opus 4.6; Calcis applies a 1.15× midpoint multiplier so your forecast already reflects that, rather than surprising you on the invoice.
Opus 4.7 also ships with a five-level thinking effort knob: low (0.5×), medium (1.0×), high (1.5×), extra-high (2.0×), and max (3.0×). The multiplier scales predicted output tokens and therefore cost. For most prompts medium is the right default; reach for high or above only when the task needs genuine chain-of-thought.
Calcis counts input tokens against Anthropic's own countTokens API - the same boundary Anthropic uses for billing - and predicts response length from the prompt before you send. Flip the thinking effort selector in the estimator to see how each level changes cost.
Estimate your cost on Claude Opus 4.7
Paste your prompt into the estimator, pick Claude Opus 4.7, and see the exact dollar cost - input tokens counted with the provider's own tokenizer, output tokens predicted by our regression model.
Frequently asked
- How much does Claude Opus 4.7 cost per request?
- At medium effort a 1,000-token prompt with a typical 500-token reply runs about $0.02 ($0.005 input + $0.0125 output, multiplied by the 1.15× tokenizer factor for Opus 4.7). Higher effort levels scale the output side: high is 1.5× the medium figure, max is 3×.
- Is Opus 4.7 more expensive than Opus 4.6?
- The rate card is identical - $5/$25 per 1M tokens. In practice Opus 4.7 bills ~15% more on average because its retrained tokenizer produces more tokens for the same text. Calcis bakes that 1.15× midpoint into the forecast.
- What do the five effort levels do?
- They set the extended-thinking budget. Multipliers on output tokens: low 0.5×, medium 1.0×, high 1.5×, extra-high 2.0×, max 3.0×. Pick the lowest level that still gets the task done.
- Does Opus 4.7 have a long-context surcharge?
- No. Anthropic bills at flat per-token rates across the full 1M context window, same as Opus 4.6.
- What tokenizer does Claude Opus 4.7 use?
- Anthropic ships a retrained tokenizer with Opus 4.7 but doesn't publish it. Calcis uses Anthropic's own countTokens API and then applies a 1.15× multiplier to match the observed 1.0-1.35× expansion vs Opus 4.6.
Pricing verified 2026-04-17 from the provider's rate card.