Two mysterious AI models just appeared on OpenRouter, and the AI community is buzzing. The Sherlock Alpha and Sherlock Think Alpha models have stealth-launched, and all signs point to xAI's next-generation Grok 4.20.

What We Know About Sherlock Alpha

OpenRouter users discovered these new models without any official announcement:

• Sherlock Alpha - Standard reasoning model
• Sherlock Think Alpha - Extended reasoning variant

Early testing reveals these aren't breakthrough frontier models, but they show meaningful improvements over existing Grok iterations:

✅ Better performance than Grok 4
✅ Faster than Grok 4 Fast
✅ Incremental capability improvements

The Grok 4.20 Connection

Multiple indicators suggest these Sherlock models are xAI's Grok 4.20:

1. Timing - Coincides with expected Grok release window
2. Performance profile - Matches anticipated incremental improvements
3. Naming pattern - "Stealth" launch aligns with xAI's testing approach

Frontier Model or Incremental Update?

With Google's Gemini 3 launching next week, Sherlock Alpha faces tough timing. If these models are indeed Grok 4.20, they may not qualify as a true frontier release—though benchmark scores could tell a different story.

The AI community will be watching closely as more testing data emerges.

What This Means for Developers

OpenRouter users can now access these models for testing. Early adopters should:

• Compare performance against Grok 4 baseline
• Test reasoning capabilities with complex prompts
• Monitor pricing and rate limits
• Share benchmark results with the community

As the AI landscape evolves rapidly, incremental improvements like Sherlock Alpha keep the competitive pressure high—even if they're not generation-defining leaps.

Keywords: Sherlock Alpha, OpenRouter, Grok 4.20, xAI, Sherlock Think Alpha, AI models, frontier models, Gemini 3, AI benchmarks