// blog · analysis · frontier-models2026-05-235 min read

The quarterly revenue doubling — Opus 4.7 + the Gemini 2.5 Pro price cut bracket the new frontier-model economics

Anthropic shipped Opus 4.7 with agent-native capabilities the same week Google cut Gemini 2.5 Pro pricing by roughly 40 percent. The two releases bracket the new frontier-model economics: capability inflection on one side, price compression on the other, and the per-task revenue math doubling quarterly. The buyer's question is no longer which model wins, it is which pricing model wins.

What landed

Opus 4.7 launched with agent-native post-training — long-horizon planning and tool use built into the model rather than wrapped around it. Google cut Gemini 2.5 Pro pricing by approximately 40 percent the same week, dropping the input-token rate below where Claude Opus and GPT-5.5 currently sit.

The doubling math

Frontier-model labs have been roughly doubling quarterly revenue through 2025 and into 2026. Anthropic's $30B run-rate this quarter, OpenAI's $122B raise valuation, the Gemini Spark agent layer launch — the throughput is real. The price cut Google just made signals confidence that the price-elasticity is high enough to grow revenue faster than the per-token margin compresses.

Capability + price compression together

The interesting tension: Opus 4.7 is positioned at the premium end with agent-native capabilities that justify higher per-task billing. Gemini 2.5 Pro just got cheaper by 40 percent at roughly comparable capability on most benchmarks. Both can grow revenue at the same time — capability inflection wins enterprise contracts, price compression wins API-developer volume. The frontier market is splitting into two layers, and the labs that win both layers grow fastest.

The frontier-model market is no longer one market. Premium-agentic and commodity-token-stream are now distinct pricing categories. Whoever ships convincingly in both grows fastest.

What the buyer signal says

Enterprise IT teams are asking for SLA-grade reliability + audit-trail + agent runtime — those are Opus 4.7 territory. API developers are asking for cheap, fast, good-enough — that's Gemini 2.5 Pro post-cut. The labs that try to compete in both with one product lose to the labs that segment cleanly. Google's price cut isn't a retreat from premium, it's a deliberate split: Gemini 3 Ultra holds the premium line; 2.5 Pro takes the volume.

The forward read

  1. OpenAI announces a parallel price split by Q3. GPT-5.5 holds premium pricing; a 'flash' or 'pro-mini' tier takes the volume at Gemini 2.5 Pro rates.
  2. Anthropic resists the price cut. Opus 4.7's agent-native capabilities are the moat; they hold premium pricing and let the volume tier go.
  3. Q3 revenue numbers double again. If the labs hit the same growth curve, the next quarterly comp is the canary for whether the doubling continues into 2027.

Anthropic — Opus 4.7 launch post → · Google — Gemini 2.5 Pro pricing update → · The Information — Frontier-model revenue tracker Q2 2026 →