The price goes up in January

Google discounting through year end, Anthropic cancelling a rise, DeepSeek raising, OpenAI cutting while selling a premium speed tier. There is no inference price trend. There is a competitive position.

Gemini 3.7 Flash costs $0.75/$3.75 per million tokens through 2026 and $1.50/$7.50 from 1 January — a doubling, announced at launch.

Do the arithmetic now

A workload at $10,000 a month becomes $20,000 on 1 January with no change in usage. For coding and agent workloads — the ones this model targets — consumption tends to grow with adoption, so the real increase is larger than the multiplier.

Announcing it up front is more honest than the alternative and produces the same outcome: teams size spend against a price with an expiry date, and the expiry lands after the architecture is committed.

Four directions in one week

Google cutting for four months. Anthropic cancelling an announced September rise. DeepSeek raising. OpenAI cutting while previewing a premium speed tier. Anyone extrapolating one direction is choosing which evidence to ignore.

What explains the spread is position, not cost. Labs fighting for a workload cut; labs that have won one raise.

The defence is architectural

Treat price as a competitive variable that can move against you, not as a technology curve that only falls. That means keeping model choice a runtime decision rather than a design commitment, so the cheapest adequate model can change month to month without a rewrite.

And remember the precedent: Anthropic published a September increase and withdrew it under pressure. Whether Google's January rise survives whatever ships in December is genuinely open.

Budget as though it holds. Be pleased if it does not.

Google — Gemini 3.7 Flash: our most intelligent workhorse model → · VentureBeat — Google's Gemini 3.7 Flash targets coding and agents with a 50% introductory price cut → · Anthropic — Pricing — Claude Platform Docs →