What was claimed

A mysterious new AI model called “Ox Alpha” is beating Claude Fable 5 and GPT-5.6 at coding with a 1 million-token context window, multimodal capabilities, and no disclosed owner (possibly from Google or Chinese lab).

Our verdict

Accurate

Claude Fable 5 is an officially announced Anthropic model, documented on Anthropic’s site and technical writeups, with a 1M-token context window, pricing details, and availability across multiple platforms.

0 of 3 AI systems agree15 sources citedChecked Aug 23, 2026

Check your own claim

Paste any statement, headline, or AI answer — 3 independent AIs verify it in seconds, with sources.

Key findings

Ox Alpha is possibly from Google or a Chinese lab

Can’t verify50%
2 of 3 AIs agree·Claude: Misleading

Ox Alpha is beating Claude Fable 5 and GPT-5.6 at coding

Verified85%
1 of 2 AIs agree·ChatGPT: Misleading

Ox Alpha has no disclosed owner

Verified88%
2 of 3 AIs agree·Claude: Misleading

Ox Alpha has multimodal capabilities

Verified92%
All 3 AIs agree

Ox Alpha has a 1 million-token context window

Verified93%
All 3 AIs agree

A mysterious new AI model called 'Ox Alpha' is beating Claude Fable 5 and GPT-5.6 at coding

Verified97%
3 of 4 AIs agree·Perplexity: Misleading

Detailed Analysis

Some elements of the claim are accurate and match current public information, but several parts are either unverified comparisons or speculative. Performance claims versus named models and the suggestion of specific possible owners are not backed by authoritative sources.

Why this verdict

  • Some elements of the claim are accurate and match current public information, but several parts are either unverified comparisons or speculative.
  • Performance claims versus named models and the suggestion of specific possible owners are not backed by authoritative sources.

Claims checked

A mysterious new AI model called 'Ox Alpha' is beating Claude Fable 5 and GPT-5.6 at coding

Misleading78%
Ox Alpha is described in its own site and OpenRouter materials as a frontier reasoning model built for coding and long-horizon agents, and it is indeed a new and somewhat stealth model.[1] [9] [12] However, none of the available sources provide rigorous benchmark evidence that it is "beating" Claude Fable 5 or GPT-5.6 at coding; performance comparisons in official materials instead emphasize GPT-5.6 Sol outperforming Claude Fable 5 on coding indices.[6] [8] [14]

This claim may mislead readers without added context.

Ox Alpha is possibly from Google or a Chinese lab

Can’t verify60%
Speculation about Ox Alpha’s origin is not supported by concrete evidence in available documentation or announcements.[9] [12] There are no credible, attributable sources confirming that Google or any specific Chinese lab is behind the model.

Ox Alpha has multimodal capabilities

Verified90%
Ox Alpha is described as accepting multiple input modalities, including text, image, and video, with browser demos supporting text plus image URLs.[12] That fits the standard meaning of multimodal capabilities.

What other AI think

The claim contains a mix of accurate and misleading elements. Ox Alpha does exist and has beaten Claude Fable 5 and GPT-5.6 on a specific coding benchmark, with a 1M-token context window and multimodal capabilities. However, the framing as "mysterious" and "no disclosed owner" is misleading—the owner is simply unlisted as "Stealth," and community analysis suggests likely origins. The benchmark results are also based on a small sample (10 tasks) with acknowledged variance.

Why this verdict

  • The claim contains a mix of accurate and misleading elements.
  • Ox Alpha does exist and has beaten Claude Fable 5 and GPT-5.6 on a specific coding benchmark, with a 1M-token context window and multimodal capabilities.
  • However, the framing as "mysterious" and "no disclosed owner" is misleading—the owner is simply unlisted as "Stealth," and community analysis suggests likely origins.
  • The benchmark results are also based on a small sample (10 tasks) with acknowledged variance.

Claims checked

Possibly from Google or Chinese lab

Misleading85%
Community analysis points specifically to Zhipu AI (Chinese lab), not Google. No credible evidence suggests Google origin. The 'possibly' hedge is appropriate, but the framing omits the leading hypothesis.

This claim may mislead readers without added context.

Ox Alpha has no disclosed owner

Misleading90%
The owner is listed as 'Stealth' on OpenRouter, not truly 'undisclosed.' Community analysis strongly suggests it's from Zhipu AI (GLM family), based on technical signatures like tokenizer and video encoder matching.

This claim may mislead readers without added context.

Ox Alpha is beating Claude Fable 5 and GPT-5.6 at coding

Verified85%
On the DeepSWE benchmark, Ox Alpha scored 80% versus Fable 5 at 65% and GPT-5.6 Sol at 52%. However, this is based on only 10 test tasks with acknowledged variance, not a comprehensive benchmark.
Some elements of the statement are corroborated by multiple community reports and listings. However the performance claim is based on early, community-run benchmarks and the origin attribution is speculative.

Why this verdict

  • Some elements of the statement are corroborated by multiple community reports and listings.
  • However the performance claim is based on early, community-run benchmarks and the origin attribution is speculative.

Claims checked

Ox Alpha is beating Claude Fable 5 and GPT-5.6 at coding

Misleading65%
Community benchmarks (DeepSWE and user-run tests) report Ox Alpha outperforming specific builds of Fable 5 and GPT-5.6 on some coding tasks, but these are early, informal, and not authoritative across all coding benchmarks.

This claim may mislead readers without added context.

Ox Alpha is possibly from Google or a Chinese lab

Can’t verify40%
Community fingerprinting and speculation suggest similarities to some known model families, but there is no verifiable evidence publicly attributing Ox Alpha to Google or any specific Chinese lab.

Ox Alpha has no disclosed owner

Verified85%
The model was released as a 'stealth' or anonymous model on platforms like OpenRouter/OpenCode with no clear attribution from a named company.

Share this result