AI Pulse · Frontier Labs Watch

Anthropic just released Opus 5 at the same price as the model it replaces, Opus 4.8, which means $5 per million input tokens and $25 per million output. That sounds unremarkable until you realise Opus 5 now outperforms Anthropic’s own premium Fable 5 model on the benchmarks that matter. It tops the Artificial Analysis Intelligence Index and posted the highest GDPval-AA v2 score according to Nous Research, per the announcement. Fable 5 costs roughly twice as much.

This is not about incremental improvement. Anthropic is collapsing its own pricing tiers to hold the middle ground before OpenAI or Google move. If you run a model that costs half as much and beats your flagship on public leaderboards, you are making a defensive play. The bet is that most enterprise buyers will not pay double for Fable 5 when Opus 5 delivers better results at the old Opus 4.8 rate. It also means GPT-5.6 and whatever Google ships next will face a lower price ceiling than they expected.

For anyone running AI workloads at scale, this changes the math immediately. If you were spec’ing Fable 5 for production, you now have a faster, cheaper alternative from the same vendor. If you were comparing Opus 4.8 to GPT or Gemini, you are comparing a materially stronger model at the same cost. The kind of command centre we build into Omni tracks these shifts in real time so you are not caught repricing contracts or rewriting prompts after a surprise release. The window between “new model drops” and “your competitor is using it” is narrowing every quarter, and price compression like this accelerates that cycle.