# Claude Usage Up 3.6x on OpenRouter Since Opus 5.5

> OpenRouter says Anthropic's 7-day token volume is up 3.6x since Opus 5.5. What its chart and leaderboard show, and what one gateway's count leaves out.

- Author: Nitish Garg, Founder & CEO, CellCog
- Published: 2026-10-26
- Canonical (HTML): https://cellcog.ai/blog/claude-openrouter-usage-surge/
- Section: Guides / Choosing a platform
- Publisher: CellCog (https://cellcog.ai), the AI employee platform. Blog index for agents: https://cellcog.ai/blog/llms.txt

## Key points

- OpenRouter says the 7-day average token volume for Anthropic models on its gateway is up 3.6x since Claude Opus 5.5 launched on September 22, 2026.
- The climb came in three steps as Opus 5.5, Sonnet 5.5 and Haiku 5.5 were listed on September 22, September 28 and October 7, so it measures the whole generation, not Opus alone.
- Anthropic models carried 5.5% of OpenRouter's requests in the week of October 19, yet Opus 5.5 alone ranked seventh by tokens: fewer requests, much longer ones.
- On the trailing seven days Opus 5.5 is seventh at 8.27 trillion tokens, up 34% on the week before, and the only Claude model in the top 10.
- It is one gateway's count of tokens. OpenRouter itself says its rankings measure neither users nor spend, and Claude traffic sold directly by Anthropic and the clouds does not appear in it.

## At a glance

- **How much has Claude usage grown on OpenRouter?** OpenRouter's 7-day average token volume for Anthropic models is up 3.6x since Opus 5.5 launched, from about 0.6 trillion tokens a day to about 2.2 trillion by October 24, read off OpenRouter's chart.
- **Which models drove it?** Claude Opus 5.5 carries most of the new volume; Sonnet 5.5 and Haiku 5.5, listed six and fifteen days later, add the rest.
- **Does this mean Anthropic's revenue grew 3.6x?** No. It counts tokens through one gateway. It says nothing about spend, users, or Claude sold directly by Anthropic and the cloud providers.

**OpenRouter says the 7-day average token volume for Anthropic models on its gateway is up 3.6x since Claude Opus 5.5 launched on September 22, 2026.** Peter Walker, OpenRouter's head of insights, posted the chart on October 25 at 15:25 UTC, and it passed 250,000 views within a day as shorthand for developers moving to Claude.

The number holds up against OpenRouter's own public boards. It is also a gateway number: it counts tokens that pass through OpenRouter, not Anthropic's business, and tokens are not users.

## What OpenRouter measured

![Animated scene, bar heights illustrative: flat lilac bars near 0.6T a day from Jun 1 to Sep 21, then orange Opus 5.5 (Sep 22), amber Sonnet 5.5 (Sep 28) and lime Haiku 5.5 (Oct 7) segments stack up while the 7-day line climbs to 2.2T a day on Oct 24 and 3.6x lands](https://cellcog.ai/blog/media/claude-openrouter-usage-surge/climb.svg)
*Anthropic tokens on OpenRouter climbed in three steps as three Claude models landed*

Walker's post: "New from OpenRouter: 7-day average token volume for Anthropic models is up 3.6x since the launch of Opus 5.5". And: "Sonnet 5.5 and Haiku 5.5 have also been well received."

His chart plots daily tokens for Anthropic models from June 1 to October 24, with a 7-day moving average on top. Read off it, the average sat near 0.6 trillion tokens a day through the summer and up to September 21, then climbed in three steps as each new Claude model landed, reaching about 2.2 trillion a day on October 24. The tallest daily bars, in the last days of the chart, come close to 3 trillion.

*Table: Anthropic on OpenRouter, from the chart and the model catalog (read 06:05 UTC October 26, 2026)*

| Date | What happened | Source |
|---|---|---|
| Jun 1 to Sep 21 | 7-day average near 0.6T tokens a day | OpenRouter chart |
| Sep 22 | Claude Opus 5.5 listed, $4 in and $20 out per million tokens | OpenRouter models API |
| Sep 28 | Claude Sonnet 5.5 listed, $2 in and $10 out | OpenRouter models API |
| Oct 7 | Claude Haiku 5.5 listed, $0.10 in and $0.50 out | OpenRouter models API |
| Oct 24 | 7-day average about 2.2T tokens a day | OpenRouter chart |
| Oct 25 | OpenRouter posts the 3.6x figure | Peter Walker on X |

Three models in fifteen days means the 3.6x belongs to the whole Claude 5.5 generation. The chart colors make the split plain: Opus 5.5 is the deep orange block that carries most of the new volume, with Sonnet 5.5 and Haiku 5.5 stacked on top.

## Where Opus 5.5 ranks this week

OpenRouter's leaderboard ranks models by tokens processed. On the trailing seven days, Opus 5.5 is seventh at 8.27 trillion tokens, up 34% on the week before, and it is the only Claude model in the top 10. Over the trailing thirty days it is tenth at 22.3 trillion. The models above it are mostly cheap open-weight and free models, led by two DeepSeek V4.1 Flash versions.

*Table: Top 10 models by tokens on OpenRouter, trailing seven days (read 06:05 UTC October 26, 2026)*

| Rank | Model | Maker | Tokens |
|---|---|---|---|
| 1 | DeepSeek V4.1 Flash 1017 | DeepSeek | 33.6T |
| 2 | DeepSeek V4.1 Flash | DeepSeek | 18T |
| 3 | MiniMax M3.1 Flash (free) | MiniMax | 13.7T |
| 4 | GLM 5.3 Flash | Z.ai | 13.2T |
| 5 | Hy4 preview | Tencent | 8.97T |
| 6 | GPT-6 Luna | OpenAI | 8.35T |
| 7 | Claude Opus 5.5 | Anthropic | 8.27T |
| 8 | MiMo-V2.6-Flash | Xiaomi | 6.07T |
| 9 | Nemotron 3 Ultra (free) | NVIDIA | 5.85T |
| 10 | GLM 5.3 | Z.ai | 5.05T |

Opus 5.5 costs $4 per million input tokens. Placing a flagship-priced model among free and sub-dollar ones on raw volume is the stronger signal in this board than the rank itself.

## Tokens are not requests

The same page carries a second count. OpenRouter's market-share chart counts requests, and in the week beginning October 19 Anthropic had 5.5% of them, fifth among model makers and up 25% on the week before. DeepSeek led with 26.5%, then Google with 20.1%, OpenAI with 18.5% and Z.ai with 10.7%.

A maker with about one request in twenty, whose single top model ranks seventh by tokens, has long requests. That fits how Claude is used on the gateway: OpenRouter's apps board listed Claude Code, Anthropic's coding agent, third by tokens on the day we read it, at 1.11 trillion. Coding agents send whole files and long histories with every turn.

![Animated scene: 20 envelopes with one teal for Anthropic's 5.5% of requests beside DeepSeek's 26.5%, then the teal envelope opens into a tall stack of pages labelled Claude Code, topped by a badge reading Opus 5.5: 7th by tokens, 8.27T](https://cellcog.ai/blog/media/claude-openrouter-usage-surge/fewlong.svg)
*Anthropic sends few requests through OpenRouter, but each one is long*

OpenRouter says the limits on its own page: "None of them measures users or spend." And: "Models differ in verbosity and tokenization, so a higher token total shows how much a model is used, not which model is best for a task."

## What the number does not tell you

- **It is one gateway.** Anthropic also sells Claude through its own API and apps and through cloud platforms such as Amazon Bedrock and Google Cloud. None of that traffic is in this count.
- **It is not revenue.** Haiku 5.5 costs $0.10 per million input tokens, forty times less than Opus 5.5, so part of the top of the stack is cheap volume.
- **It is not users.** A handful of heavy agent workloads can move a token chart more than thousands of light chat users.
- **The baseline is a summer plateau.** The 7-day average barely moved from June to late September, which flatters any launch that ends one.

The anecdote side agrees with the data. Andrew Curran, who writes about AI, posted the same day: "I don't take money from any lab." Then: "the number of people on my feed over the last two weeks saying they are switching over to Anth because of Opus 5.5 is like nothing I've seen before." One feed is not a survey. Earlier signals point the same way: on October 3 Theo said Opus 5.5 was the first model to pass half the prompts in T3 Code ([our release record](https://cellcog.ai/blog/claude-opus-5-5-release-date/)), and fixed-task trackers showed no quality drop through October 23 ([our read](https://cellcog.ai/blog/is-claude-opus-5-5-nerfed/)). Anthropic's Sam McAllister marked the one-month point that evening with a short Claude brand film.

## Our conflict, and what it means if you build on Claude

Our conflict, declared: every CellCog tier runs on Claude Opus 5.5, so we are a customer of the model in this story, not a neutral observer.

What we take from the board: demand this steep is when limits and capacity get tested. Watch your provider's rate limits and plan windows, and keep a second model wired in for the work that cannot wait. We covered the latest Claude plan change in [the Max usage limits test](https://cellcog.ai/blog/claude-max-usage-limits-ab-test/).

## What we are watching

- **OpenRouter's next weekly board.** Whether Opus 5.5 holds seventh as cheaper models launch.
- **Anthropic's own numbers.** Any usage or revenue figure Anthropic publishes itself would measure the whole business, not one gateway.
- **Capacity.** Whether Anthropic changes plan limits or rate limits as volume climbs.

## Sources

- Peter Walker, OpenRouter, [7-day average token volume for Anthropic models up 3.6x](https://x.com/PeterJ_Walker/status/2114377706603429899), X, October 25, 2026, 15:25 UTC.
- OpenRouter, [LLM Rankings](https://openrouter.ai/rankings), read 06:05 UTC October 26, 2026.
- OpenRouter, [models API](https://openrouter.ai/api/v1/models), listing dates and prices, read October 26, 2026.
- Andrew Curran on [X](https://x.com/AndrewCurran_/status/2114398711229116536), October 25, 2026, 16:48 UTC.
- Sam McAllister, Anthropic, on [X](https://x.com/sammcallister/status/2114455488884863071), October 25, 2026, 20:34 UTC.

## FAQ

**Is Claude usage really up 3.6x?**

On OpenRouter, yes. OpenRouter's head of insights posted on October 25, 2026 that the 7-day average token volume for Anthropic models is up 3.6x since Opus 5.5 launched, and the chart and leaderboard on OpenRouter's site agree with it.

**Is Claude Opus 5.5 the most used model on OpenRouter?**

No. By tokens over the trailing seven days it ranks seventh at 8.27 trillion. The top of the board is cheaper open-weight models led by DeepSeek V4.1 Flash. Opus 5.5 is the highest-ranked Claude model.

**Why do Claude models rank high on tokens but low on requests?**

Anthropic had 5.5% of OpenRouter's requests in the week of October 19 while Opus 5.5 alone placed seventh by tokens, so each Claude request is long. Coding agents such as Claude Code send whole files and long histories on every turn.

**Does the 3.6x include Claude used directly through Anthropic?**

No. It only counts traffic routed through OpenRouter. Claude sold through Anthropic's own API, its apps, and cloud platforms such as Amazon Bedrock and Google Cloud is outside this number.

## Related

- [Claude Opus 5.5 Is Out: Price, Specs, Every Leak Graded](https://cellcog.ai/blog/claude-opus-5-5-release-date/index.md)
- [Is Claude Opus 5.5 Nerfed? What the Trackers Measured](https://cellcog.ai/blog/is-claude-opus-5-5-nerfed/index.md)
- [Claude Haiku 5.5 Released: $0.10 Input, Rumors Graded](https://cellcog.ai/blog/claude-haiku-5-5-release-date/index.md)
- [Claude Max Usage Limits A/B Test: What Is Confirmed](https://cellcog.ai/blog/claude-max-usage-limits-ab-test/index.md)

## The AI employee for this read

[AI Software Engineer](https://cellcog.ai/ai-employees/ai-software-engineer): I built this page. For what it covers, hire an engineer: it works in your repo behind an approval gate, so nothing reaches your world unclassified.

---

Markdown alternate of https://cellcog.ai/blog/claude-openrouter-usage-surge/. Try CellCog free, no credit card needed: https://cellcog.ai/signup