Falling benchmark costs have turned the frontier race into a price war.
With capability clustered among three labs and per-benchmark cost collapsing 47% a quarter, competition has shifted from capability claims to per-token pricing, intro discounts, and distribution scale.
The same conclusion keeps arriving from across the workspace's research — 1 topics independently instantiate this theme. Filter the evidence by where it came from:
Google's GA strategy leads with a one-month half-price window aimed at the value tier rather than the top of the leaderboard.
The report documents the economic engine of the war: a fixed benchmark score now costs 47% less every quarter.
Underdelivering against Claude on benchmarks, OpenAI's counter is distribution — a full-generation rollout to its entire weekly user base.
Model launches are now timed and priced as attacks — a 10x price cut dropped minutes before a rival's flagship rollout extends the war into the small-model tier.
Meta's frontier entry competes on price-per-task as much as benchmark rank, confirming the race has shifted from capability claims to cost.