Model pricing
Claude 3 Sonnet pricing
Claude 3 Sonnet by Anthropic costs $3 per 1M input tokens and
Claude 3 Sonnet pricing
5 per 1M output tokens (cached input $0.3, batch input not recorded) as of 2026-10-31. A 1,000-input/500-output request costs $0.0105. The lowest-cost alternative with at least this context window and the same recorded reasoning category is Ling 2.6 Flash at $0.01/$0.03; capability equivalence is not guaranteed.Notes
Original Claude 3 Sonnet — balanced reasoning and cost
Key pricing facts
| Fact | Value |
|---|---|
| Input / 1M tokens | $3 |
| Output / 1M tokens | $15 |
| Cached input / 1M | $0.3 |
| Batch input / output / 1M | Not recorded |
| Context window | 200,000 |
| Provider | Anthropic |
| Last verified | No model-specific verification date recorded |
| Snapshot updated | 2026-10-31 |
Last updated: · Source: Anthropic pricing · Methodology
Frequently asked questions
What does Claude 3 Sonnet cost per million tokens?
Input costs $3 and output costs $15 per million tokens. The 1,000-input/500-output example costs $0.0105, before discounts, taxes, hidden reasoning tokens or context-tier surcharges.
Does Claude 3 Sonnet have cached or batch pricing?
Cached input: $0.3 per million tokens under the existing provider cache rules. Batch: not recorded in this dataset. Eligibility and cache-write charges depend on the provider; these discounts are not automatically stacked.
What is a cheaper alternative to Claude 3 Sonnet?
Ling 2.6 Flash costs $0.01 input and $0.03 output per million tokens. This comparison matches context capacity and reasoning category, not benchmark quality. Test your workload before replacing a model.
How often are these prices verified?
The live pricing table synchronizes stored model records daily; public file snapshots update when the site is rebuilt. Synchronization is not independent provider verification. Snapshot updated 2026-10-31. No model-specific verification date is recorded; the snapshot date must not be interpreted as an official provider verification.