ActiveSpans 6 functions & 5 industries
Updated

Claude Opus 4.8 Rockets Past GPT-5.5—But Soaring Token Costs Ignite AI Arms Race Backlash

AI models are getting smarter by splitting work into swarms — and the bill is becoming the battleground.

What is this trend?

Multi-agent AI workflows are pushing frontier models to tackle bigger tasks faster, but rising token and compute costs are forcing buyers and vendors to rethink what “better” really means.

  • Parallel subagents can compress complex work from weeks to days, especially in coding and enterprise automation.
  • Raw benchmark wins matter less when real-world gains are uneven and evaluation methods look saturated.
  • Token-heavy orchestration can erase speed gains by driving API and compute costs sharply higher.
  • Cheaper near-frontier rivals are turning pricing into a core competitive weapon.
  • The race is shifting from model size alone to efficiency, transparency, and sustainable deployment.

What’s the latest?

Anthropic’s Claude Opus 4.8 dazzles with dynamic, multi-agent workflows and benchmark wins, but high token costs and 'meh' code leave users divided.

How it developed earlier updates

  1. Anthropic’s Claude Opus 4.8 has outgunned GPT-5.5 in coding benchmarks, but its sky-high token costs are stoking a fierce AI arms race and backlash over unsustainable pricing.

    AI Price War Heats Up as Enterprises Demand 90% Cut

Where this is playing out

Related trends

Stay ahead of what’s changing

Get the weekly brief and deep-dive reporting in your inbox.