Anthropic
Claude Sonnet 4.5 pricing
Previous-generation Sonnet. 5x cheaper than Opus on the same tokenizer, strong enough for most production workloads.
Input
$3.00/ 1M tok
Output
$15.00/ 1M tok
- Context window
- 200K
- Max output
- 64K
- Cached input
- -
- Verified
- 2026-04-06
Sonnet 4.5 is the direct predecessor to Sonnet 4.6. Price is identical ($3 / $15 per 1M tokens), context window is the same 200K, and max output is 64K. The gap versus 4.6 is a shift in capability rather than cost - 4.6 is measurably better at complex reasoning and long-context tasks, 4.5 is still strong on most typical chat and summarisation workloads.
At 5x less than the Opus tier and with the same API surface, Sonnet 4.5 remains a sensible default for production code that was benchmarked on this version. A 10K-token prompt with a 2K-token reply comes in at about $0.06 - half the cost of Opus for the same shape of request.
Calcis counts input tokens against Anthropic's own countTokens API for an exact figure before you send, and predicts response length from the prompt rather than asking you to guess.
Estimate your cost on Claude Sonnet 4.5
Paste your prompt into the estimator, pick Claude Sonnet 4.5, and see the exact dollar cost - input tokens counted with the provider's own tokenizer, output tokens predicted by our regression model.
Frequently asked
- Should I use Sonnet 4.5 or 4.6?
- Default to 4.6 for new work - same price, newer training, better complex-reasoning performance. Stay on 4.5 only if you've pinned this version for reproducibility on an existing benchmark.
- How much does Sonnet 4.5 cost per request?
- A 1,000-token prompt with a 500-token reply costs about $0.0105 ($0.003 input + $0.0075 output). That's 5x cheaper than Opus 4.5 for the same request shape.
- Is Sonnet 4.5 cheaper than GPT-5?
- On input, no - $3 vs GPT-5's $1.25 per 1M tokens. On output, yes - Sonnet's $15 output is cheaper than some GPT-5 variants but roughly matches the full GPT-5 rate. Pick on capability and ecosystem, not just headline price.
- Does Sonnet 4.5 support prompt caching?
- Yes. Anthropic's prompt caching applies to Sonnet 4.5 the same way as other Claude models - cache writes at 1.25x input rate, cache reads at 0.1x (a 90% discount on repeated prefixes).
Pricing verified 2026-04-06 from the provider's rate card.