Claude Mythos Preview vs Llama 4 (Meta)

Tone: When Anthropic does publish Mythos outputs (in sanitized research reports), the voice is careful, technically dense, and deliberately unperformed -- much more 'senior security researcher writing an internal memo' than Claude Opus's conversational style.

Quirks: Mythos is tuned to produce its cybersecurity reasoning with extensive show-your-work traces. Anthropic publishes some outputs with full CoT visible as evidence of capability claims. Outside of security tasks, the model reportedly sounds much like Opus 4.6 / 4.7 -- Anthropic hasn't published a distinct general-purpose voice for Mythos.

Tone: Plain, helpful, and neutral. Meta's instruction-tuned Llama 4 reads like a sanitized ChatGPT -- useful for general tasks but without a strong persona of its own.

Quirks: The 'real' personality depends on the checkpoint you run. Base Llama 4 is bland by design; the interesting behaviors come from community fine-tunes (Nous, Hermes, Dolphin, etc.) that give it different voices and refusal patterns.