Anthropic Token Volume on OpenRouter Is Up 3.6x Since Opus 5.5 Launch, According to a Post on X

Token volume for Anthropic's models is up 3.6x since Claude Opus 5.5 went live on September 22, 2026, according to a post on X by user @peterj_walker. The figure is a 7-day average, and it describes traffic on OpenRouter, a platform that routes developers' requests to AI models.

That is not a company-wide number. Anthropic has not published a total usage figure for the period, and the poster has not shown the underlying data.

What Anthropic changed with Opus 5.5

Opus 5.5 is the first model in the Claude 5.5 family. Anthropic aimed it at long-running coding, research and professional work, the multi-step jobs where an AI agent works through a task rather than answering one question.

The pricing is where the model stands apart. Opus 5.5 costs $4 per million input tokens and $20 per million output tokens, a 20% cut from Opus 5.

Cache reads fell 60%, to $0.20 per million tokens. Caching lets a model reuse text it has already processed, such as a large codebase, so agents that keep returning to the same material pay far less.

Anthropic says overall operating costs are down about 40% compared with Opus 5. Output generation is more than 30% faster. Usage limits on Pro, Max and Team plans rose 20%, and the model has a new, less verbose communication style.

On performance, Anthropic says Opus 5.5 performs at the level of Claude Fable 5.1 on most work. Benchmark results show it matching or beating the larger Fable 5.1 on tests including FrontierCode and CursorBench.

Those are the company's figures. No independent validation of the 40% cost claim has surfaced yet, and enterprise buyers moving production traffic will want it.

Fewer tokens per task, more tasks overall

The numbers look like they conflict. Box, one of the early testers, reported cutting token consumption by one-third while keeping accuracy intact. Yet aggregate volume on OpenRouter jumped 3.6x.

If each job uses fewer tokens, total volume only climbs that fast when far more jobs are being run. Crypto Briefing's Diego Almada Lopez points to the Jevons paradox: when a resource gets cheaper and more efficient to use, total consumption often rises.

That is an explanation, not a finding. Nobody has published data showing why volume rose, how much of it is new customers versus existing ones running more work, or how much of the jump reflects OpenRouter's own traffic mix.

Users say they are switching

AI commentator Andrew Curran wrote on X that the number of people in his feed over the last two weeks saying they are switching to Anthropic because of Opus 5.5 is "like nothing I've seen before." That is one person's view of his own timeline, not a measure of market share.

SaaS investor Jason Lemkin wrote that the 5.5 line is "pretty darn close to Sonnet for 10%-15% of the price." Anthropic has not published a comparison framed that way.

The rest of the 5.5 lineup

Opus 5.5 was followed by Claude Sonnet 5.5 on September 28, six days later. Anthropic's release notes say Claude Haiku 5.5, the smallest and cheapest tier, is due soon. That follows the company's pattern of shipping the flagship first, then the mid-tier, then the small, fast model built for high-volume work.

The pace sits awkwardly with a public pledge. Shattered.io reports that Anthropic has committed to a slower, more deliberate release cadence under its own pacing pledge, while the 5.5 family has arrived in a matter of days. Anthropic has not publicly reconciled the two in the material cited here.

Anthropic also faces competition at the top. Comparisons with OpenAI's GPT-6 and xAI's Grok 4.7 are circulating, but the sources here offer no benchmarks that settle that contest.

The open question is whether the volume spike holds. A 7-day average taken weeks after launch can reflect a burst of testing and migration that fades once developers finish evaluating. Cheaper per-token pricing helps only if customers keep sending work.

Anthropic's Haiku 5.5 release will test the cost story. If the smallest tier lands at the price levels users are already describing, the company will have to show that revenue keeps pace with the volume it is serving at lower prices. Anthropic has not disclosed usage or revenue figures for the Claude 5.5 family.