---
format: "aidr-story-markdown/v1"
id: "ce097061bd01321f4dc23bb5a9da955b9300dd32bb8462eb81231f53f293f340"
canonical_url: "https://aidr.today/ce097061?lang=en"
title: "Ox Alpha Processes 26T Tokens in 4 Days, 2.6x OpenRouter's Biggest Launch"
lang: "en"
requested_lang: "en"
available_langs: ["en","vi"]
translation_fallback: null
fallback_fields: []
published_at: "2026-08-25T05:48:04.000Z"
category: "Models"
topics: ["agent","coding","llm","open-source","harness","inference","benchmark","deepseek"]
source_urls: ["https://huggingnews.com/ai/ox-alpha-processes-26t-tokens-in-4-days-26x-openrouters-biggest-launch-4010d65f","https://x.com/rohanpaul_ai/status/2092050483422855648","https://x.com/ycombinator/status/2092086376556863627","https://x.com/nicbstme/status/2092027775238672519","https://huggingnews.com/ai/ox-alpha-processes-26t-tokens-in-4-days-to-set-openrouter-launch-record-8eb0f8db","https://x.com/vipulved/status/2092387838566482181","https://x.com/AAAzzam/status/2092391013595677140","https://x.com/togethercompute/status/2092388843429478751"]
summary: "Coding agents leveraging repeated tool calls and long contexts drove a surge in token consumption for the newly released Ox Alpha model. The model processed 26 trillion tokens in four days on the OpenRouter platform, a volume 2.6 times larger than any previous model launch on the service. To support sustained agentic work, Ox Alpha utilizes a 1.05M token context window that allows sessions to carry more text than typical chat applications. Single-day usage for the model reached 5 trillion tokens, roughly 30 times the daily volume of Claude Opus 5 or GPT-5.6 Sol. The model is currently offered for free via the ori harness to encourage adoption among developers using coding agents."
---

# Ox Alpha Processes 26T Tokens in 4 Days, 2\.6x OpenRouter's Biggest Launch

> [Open the canonical story](<https://aidr.today/ce097061?lang=en>)

**Published:** 2026-08-25T05:48:04.000Z
**Category:** Models
**Topics:** agent, coding, llm, open\-source, harness, inference, benchmark, deepseek

## Summary

Coding agents leveraging repeated tool calls and long contexts drove a surge in token consumption for the newly released Ox Alpha model\. The model processed 26 trillion tokens in four days on the OpenRouter platform, a volume 2\.6 times larger than any previous model launch on the service\. To support sustained agentic work, Ox Alpha utilizes a 1\.05M token context window that allows sessions to carry more text than typical chat applications\. Single\-day usage for the model reached 5 trillion tokens, roughly 30 times the daily volume of Claude Opus 5 or GPT\-5\.6 Sol\. The model is currently offered for free via the ori harness to encourage adoption among developers using coding agents\.

## Sources

- [Story source](<https://huggingnews.com/ai/ox-alpha-processes-26t-tokens-in-4-days-26x-openrouters-biggest-launch-4010d65f>)
- [Supporting source](<https://x.com/rohanpaul_ai/status/2092050483422855648>)
- [Supporting source](<https://x.com/ycombinator/status/2092086376556863627>)
- [Supporting source](<https://x.com/nicbstme/status/2092027775238672519>)
- [Story source](<https://huggingnews.com/ai/ox-alpha-processes-26t-tokens-in-4-days-to-set-openrouter-launch-record-8eb0f8db>)
- [Story source](<https://x.com/vipulved/status/2092387838566482181>)
- [Story source](<https://x.com/AAAzzam/status/2092391013595677140>)
- [Supporting source](<https://x.com/togethercompute/status/2092388843429478751>)