| Claude Sonnet 5 | Claude Sonnet 4.6 | |
|---|---|---|
| Price in/out ($/M) | $2 / $10 | $3 / $15 |
The rare upgrade that costs less: Sonnet 5 lands at $2 / $10 per M tokens against $3 / $15 for Sonnet 4.6, and takes the context window from 200K to 1M. For production coding copilots and agents that have to justify their cost, there is no argument for staying on 4.6. The one thing to instrument first is the new adaptive effort selector, which makes spend track task difficulty and therefore makes per-call cost harder to forecast than 4.6's flat behaviour. Opus still leads on the hardest tasks.
Claude Sonnet 5 · Claude Sonnet 4.6 · All AI model comparisons