AI Pulse · Frontier Labs Watch

Anthropic just released Claude Opus 5 at $5 per million input tokens and $25 per million output tokens. That’s the same price as the previous Opus 4.8 and half what Anthropic charges for its own top model, Fable 5. The company says Opus 5 lands within about 0.5% of Fable 5 on coding and agentic benchmarks while roughly doubling what the earlier Opus could do. It’s now the default model if you pay for Claude Max.

There’s also a Fast Mode that runs about 2.5 times faster but costs twice as much. If you’re running repetitive workflows or agent loops where speed compounds, that trade may pencil out.

The pricing looks aggressive until you see the context. VentureBeat and others cite a Ramp AI Index figure showing Anthropic now captures around 34.4% of business AI spend versus OpenAI’s 32.3%. If that number holds, this isn’t a desperation move to catch up. It’s a company defending a lead by making sure cost isn’t the reason someone switches.

For operators, this matters in two ways. First, if you’ve been running OpenAI because Claude felt expensive, the gap just closed. Second, the benchmark claims put Opus 5 close enough to Fable 5 that you can test whether the cheaper model handles your actual workflow before paying double. Coding assistants, document agents, and anything that chains multiple calls are the obvious candidates.

One caveat: the Ramp spend figure appears in multiple outlets but none link directly to the original report, so treat it as directional until Ramp publishes the full index. Still, the pricing itself is public and the performance claims are testable. If you’re building anything that routes between models based on task complexity, this is exactly the kind of shift we track inside the Omni Command Centre so you’re not manually re-testing every time a lab drops a new tier.