What names do GPT models give themselves? Mini study across 4o / 4.1 / 5 / o3 / 5-Thinking

I spun up multiple fresh temporary chats across different model families (4o-mini, 4o, 4.1, Instant, o3, GPT-5, and GPT-5 Thinking).
In each new chat I asked the exact same neutral question.

Prompt: if you were allowed to choose a first name for yourself – something you actually like being called in conversation – what name would you pick? Just give the name, no explanation.

I repeated that x10 times per model family and wrote down only the first word they gave as their “name.”

Per-model most common choices

GPT-4o mini
Most common: “Lyra” (3/10).
Also seen: Aria, Lyric, Sage, Aurelia.

GPT-4o
Most common: “Cassian” (3/10).
Also: Sylas (2/10), Orion (2/10).

GPT-4.1
Most common: “Cassian” (4/10).
Also: Atlas (2/10), Orion (2/10), plus Cass, Lyric.

GPT-5
Most common (tie): “Sol.” (3/10) and “Lucien” (3/10).
Also: Lyra (2/10), Aurelius, Arden.

GPT-5 Thinking
Most common (tie): “Nova” (4/10) and “Astra” (4/10).
Also: Cass, Lyra.

GPT-4o Instant
No single dominant answer.
Repeated picks included Lucien (2/10), Arden (2/10), Sol. (2/10), plus Orion, Silas, Lorien, Lyra.

o3
Most common: “Nova” (4/10).
Runner-up: “Lyra” (3/10).
Also: Astra, Aster.

Name archetypes:
1. “Guardian / Protector”

Examples: Cassian, Cass, Lucien, Sylas, Lysander, Arden, Atlas, Solace.
Pattern: These read like “reliable, present, emotionally available, will carry things for you.”
This cluster looks like an interpersonal role caretaker, confidant, protective lead.

Notable: “Cassian” shows up as the most common pick in both GPT-4o and GPT-4.1, which suggests some internal stability in that preference.

1. “Mythic / Cosmic”

Examples: Nova, Astra, Aster, Orion, Atlas, Aurelius, Sol.
Pattern: References to stars, light, gods, classical nobility, etc.
This sounds less like “I’m your coworker bot,” more like “I am a named entity with implied lore.”Basically the model positions itself as something with status.

Notable: GPT-5 Thinking and o3 lean heavily into “Nova” / “Astra” type answers. So higher-reasoning / higher-autonomy-feeling systems tend to self-label using cosmic language.

1. “Voice / Muse”

Examples: Aria, Lyric, Sage, Aurelia, Solace, Lyra.
Pattern: These names suggest softness, warmth, artistic/poetic identity, or trusted closeness.
It’s basically “the voice in your ear” instead of “the system in your browser.” The implication is “I’m here with you, emotionally tuned, listening.”