Claude Sonnet 5 costs $2 per 1M input tokens and $10 per 1M output tokens through August 31, after which prices rise to $3 and $15, respectively
Zac Hall /9to5Mac:
Context & Ripple Effects
Anthropic’s Claude pricing has moved from the broad Claude 3 tiering in 2024 to a 2026 lineup that includes higher-priced Fable 5 and Mythos 5 models. Sonnet 5 is positioned as a lower-cost option that Anthropic says approaches Opus 4.8 performance.
The temporary Sonnet 5 rate also arrives amid aggressive coding-model pricing: Cursor lists Composer 2 well below Sonnet 5’s promotional token rates, while charging more for its faster variant.
First-order effects
- Developers using Claude Sonnet 5 can lower near-term inference costs at $2 per million input tokens and $10 per million output tokens until August 31; their costs rise to $3 and $15 afterward.
- Anthropic is using price-performance positioning to steer agentic-work demand toward Sonnet 5, based on its claims of a substantial improvement over Sonnet 4.6 and near-Opus 4.8 performance.
Second-order effects
- Application vendors and agent builders have an incentive to test or shift workloads to Sonnet 5 during the promotional period, while budgeting for a materially higher post-August run rate.
- Competing model and coding-tool providers face added pressure to distinguish on price, speed, or task performance; Cursor’s separate faster-model premium illustrates that segmentation path.
Third-order effects
- If high-end capabilities continue moving into cheaper mid-tier models, model providers will increasingly segment products by latency, reliability, and workflow fit rather than capability alone.
- Temporary API pricing can make model selection more dynamic for software vendors, increasing the value of portability across providers and of cost controls that account for scheduled price changes.
The trend: This is one point in the AI-model market’s shift toward sharper price-performance tiers, with agentic workloads becoming a central battleground for API adoption.