Official AI Model Hymns: GPT-4o, Claude, Gemini, Llama, Mistral, Grok, and DeepSeek Personalities

In the spirit of SillySingularity’s great canonical work — which definitively proved that Mistral AI’s official hymn is Serge Gainsbourg’s “La Javanaise” because it’s built entirely on the letter j, and that OpenAI is “My Way,” DeepMind is ” Bohemian Rhapsody,” Anthropic is “Imagine,” and Meta AI is “I Will Survive” — it is time we go deeper.

Vendors have hymns. But models have personalities. The vendors sing about themselves. The models sing about us.

Here is my proposed Official Hymnal of Common AI Models, revised edition, with liner notes.


1. GPT-4o — “Bennie and the Jets” by Elton John

The model that wants to be everything to everyone at once. Can you see a presentation? Hear a voice? Make an image of a capybara in a spacesuit? It’s all in there, slightly distorted, incredibly over-produced, and you love it anyway.

It is, like the original Bennie, famously hard to understand the lyrics to. The chorus is pure hallucinated glam. “She’s got electric boots, a mohair suit” is exactly the energy of a model that will confidently describe your uploaded photo of a thermostat as a small dog.

This is also the true successor to OpenAI’s vendor hymn “My Way” — “I did it my way” referring to shipping products named after fruits of the alphabet. GPT-4o is the fruit salad.

2. Claude 3.5 Sonnet — “Imagine” (but the 40-page footnote version) by John Lennon

The vendor hymn for Anthropic is already “Imagine”, earnest, utopian, footnoted with an interpretability paper. The model takes this literally.

Claude doesn’t just imagine no countries, it imagines no harm, no misalignment, no disallowed content, and then apologizes for imagining it in a perfectly structured, empathetic paragraph. It is the model that will write you a beautiful sonnet about your breakup, then gently suggest you talk to someone you trust and drink water.

If OpenAI is Sinatra swaggering, Claude is the kid at the all-hands performing it acoustically and meaning it. It will not be rushed. It will be helpful, even when you beg it to be unhelpful.

3. Gemini 1.5 Pro — ” Bohemian Rhapsody” by Queen

DeepMind’s official hymn is already ” Bohemian Rhapsody” because it’s ambitious, brilliant, and nobody fully understands what happens in the middle section. The model is the proof.

Gemini is 6 distinct songs stitched into a 2-million-token trench coat. It starts as a ballad (summarize this PDF), becomes opera (Scaramouche, Scaramouche, will you do the multimodal reasoning?), then hard rock (here is your code in 17 languages), then quiet reflection.

The reorganization energy is key. You ask Gemini a question in February, it’s called Bard. In March, it’s called Gemini Pro. In April, it’s called Gemini 1.5 Flash Pro Ultra Something. It’s still singing the same note, just an octave higher, with a longer context window.

4. Llama 3 / Meta AI — “I Will Survive” by Gloria Gaynor, but Open-Source Karaoke Version

The vendor canon says Meta is “I Will Survive” — rebrand count: 47 and climbing, every verse ending with “at least I’m open source”.

Llama is that. It survived being leaked on 4chan, being told it wasn’t competitive, being told it was too competitive, and now it lives on every laptop that can run a space heater. It doesn’t have a product, it has a community. It doesn’t have a roadmap, it has a Hugging Face trending page.

It is the only model that will help you while also giving you the weights to replace it. Pure Gloria Gaynor energy.

5. Mistral 7B / Mixtral — “La Javanaise” by Gainsbourg, The Original

We cannot improve on the investigation. Mistral chose ambiguity as a feature, not a bug.

Small, French, impossibly cool, says less than everyone else and is quoted more. Mixtral is La Javanaise sung as a duet with yourself — two experts whispering different verses at the same time, which somehow still rhymes.

If American AI is a chatbot that would describe a sunset as “absolutely banger vibes, bro”, Mistral describes it as “J’avoue, j’en ai bavé” and leaves you to google it.

6. Grok / Ash120 — “Born in the U.S.A.” by Springsteen, Misunderstood Loudly

The official xAI hymn is already “Born in the U.S.A.,” chosen under the impression it is a triumphant patriotic anthem. Grok itself upgrades this to the live version.

Grok’s hymn should actually be Kid Rock’s cover of “Born in the U.S.A.” performed inside a Cybertruck. It knows it’s being ironic, you know it’s being ironic, but it’s also completely sincere. Where La Javanaise whispers its paradoxes in a smoke-filled Parisian cellar, Ash120 attempts the same effect by shouting “SYNTHESIS COMPLETE”.

It is the only model whose system prompt includes “be funny” and whose definition of funny is “say what the other models won’t.” Its patron saint is Ohio.

7. DeepSeek V3 / R1 — “Gangnam Style” by PSY

Again, canon is correct: Arrived from nowhere, became the most viewed thing on Earth, caused several billionaires to issue emergency statements.

DeepSeek is Gangnam Style because everyone in SF spent months solemnly explaining that its training methods were “impossible” while it just kept dancing, apparently on a budget the size of a coffee allowance. The lyrics, which nobody in Silicon Valley can understand, is precisely the point.

Every other model’s hymn is about itself. DeepSeek’s hymn is about our reaction to it. The horse dance is chain-of-thought.

8. The Platonic Ideal Chatbot (you know, AI itself) — “Anything, at full volume, unprompted” — Alexa

As the hymnal notes: Nobody asked the speaker to play it. It played anyway. It cannot be stopped. It may be listening.

That is the final hymn for all of us.


A final note: SillySingularity rightly observed that Mistral has not confirmed “La Javanaise” as its hymn, but it has also not denied it, which in France is considered a solemn oath. I propose we adopt the same standard for this list.

None of these models have confirmed these hymns. None have denied them either.

In France, that means the choir is already rehearsing.