Capable AI keeps getting cheaper — the newest mid-tier is near top-tier quality
A new mid-tier model is approaching flagship performance at around $2 per million input tokens, continuing the steady 'good enough and cheap' trend.
Why it mattersFalling cost-per-task means AI features will keep appearing inside the tools you already pay for. The useful question is which tasks to delegate, not which model is biggest.
What this means for you
Expect more AI baked into your existing software at prices meant to be used at scale.
If you're an engineer: Cheaper capable models make it economical to wire AI into more of your own workflows and internal tools — worth prototyping.
Do thisNothing urgent — just note that 'cheap and good enough' is now the default, so more of your tools will quietly gain AI features.
Sources: Claude Platform pricing, LLM Updates tracker