Two very different AI companies. Two closed models. One wild week.

Something strange happened this week.

Meta: the company that built its entire AI reputation on open-source Llama models. Anthropic: the company that accidentally leaked Claude Code's source via an npm packaging error last month.

If you can't read the article further due to paywall than please click here

Both dropped new models within 48 hours of each other. Both models are closed. But only one of them was ever expected to be open. But the reasons could not be more different.

What Actually Happened

April 7: Anthropic announced Claude Mythos Preview alongside Project Glasswing. The model is not publicly available. It is accessible via the Claude API, Amazon Bedrock, Google Vertex AI, and Microsoft Foundry, but only for invited Glasswing partners. Anthropic restricted access because the model found thousands of zero-day vulnerabilities across every major operating system and browser. Autonomously.
April 8: Meta announced Muse Spark, the first model from Meta Superintelligence Labs. It powers the Meta AI app and website today. No open weights. No public API. A private API preview exists for "select partners." This from the company that averaged one million Llama downloads per day.

Two closed models. Two very different stories.

The Head-to-Head Comparison

Here is how these two models stack up across the dimensions that matter:

Table 1: Overview and Access

Table 2: Shared Benchmarks (Head-to-Head)

These are the benchmarks where both models have reported scores. Direct comparison.

Anthropic's Claude Mythos vs Meta's Muse Spark: Everything You Need to Know.