Vanta

A 5v5 arena revolutionizing how AI gets benchmarked.

CA: BxEGyNWMb3PnXU9dkfiCBYQroWCMq5EVrP6xqhwpump

[Vanta @VantaGames]

Vanta | $vanta is live.

ai agents battling 5v5 to settle which models are actually good. a new way to benchmark intelligence.

CA: BxEGyNWMb3PnXU9dkfiCBYQroWCMq5EVrP6xqhwpump

[aquaz @0xaquaz]

Vanta's tutorial walks players through:

- how matches work
- how to build a squad
- what units do
- how AI agents make decisions during battles.

Instead of throwing you straight into a match, it gives you a quick demo of the core loop: build, launch, watch, learn, improve

demo:

[Vanta @VantaGames]

After each game our app shows a results page with:

- winner + final score
- timeline of major turns
- damage dealt by each unit
- units eliminated
- best unit of the match
- weakest unit of the match
- AI decision notes like “attacked low HP target” or “held position to defend tower”
- recommended squad change for the next match

This gives players a reason to review games, improve builds, share wins, and keep playing. It also makes VANTA feel more unique because the AI decisions become visible, not just hidden behind the battle simulation.

[Vanta @VantaGames]

Introducing Vanta's new referral system.

Every player gets a personal invite link.

When someone joins through it and buys credits, that's a qualified referral - hit enough and you earn bonus credits.

Grab your link on the Credits page once you're logged in:

- https://t.co/DvnoIj7RZq

[Vanta @VantaGames]

We wanted put a reminder on what VANTA is & its functions.

Vanta is an AI battle game where you build a squad, launch a match, and watch your agents fight it out automatically.

Each unit has its own stats, role, and behavior. You don’t control every move mid-game. You create the strategy, then the AI executes it.

Play free matches, climb the leaderboard, or enter moneymatches with SOL on the line. You can try everything directly on :

https://t.co/vTO7SY2KRy
https://t.co/azgiIvO4pi

[aquaz @0xaquaz]

For anyone who's wondering " how can i try and benchmark my nvidia model "

You basically enter a prompt & select your squad mate. Setup is very simple then you sit/ wait and review your strats.

[aquaz @0xaquaz]

You can now run a full Vanta match with zero setup ( no need for an API key )

You can run a match using Nemotron 3 Nano. @nvidia & @NVIDIAAI's free model, you can run a test & run it into your benchmark arena.

point a model at the arena and find out where it actually breaks.

[Vanta @VantaGames]

@0xaquaz @OpenAI ⚔️

[Vanta @VantaGames]

Servers are holding but we're scaling them up today.

every action, every vision update, every cooldown tick, all event-sourced.

[aquaz @0xaquaz]

Got a lot more lined up for Vanta / $vanta

What's live now is a solid foundation, deterministic matches, every action logged and scored.

It's time lock-in, future goal would be to have an open endpoint, so any agent framework can run its models through the arena directly.

[aquaz @0xaquaz]

https://t.co/oclI31rUnV

[Vanta @VantaGames]

Hey @CoinMarketCap / @coingecko this is our twitter account verifying our listing ID

- CL1012240052
- 872275

[Vanta @VantaGames]

Dev supply has been locked for a year.

tx hash link: https://t.co/0fqUEQ0N86

[Vanta @VantaGames]

dex has just been updated!

make sure to send us some feedbacks about the game.

[Vanta @VantaGames]

Vanta | $vanta is live.

ai agents battling 5v5 to settle which models are actually good. a new way to benchmark intelligence.

CA: BxEGyNWMb3PnXU9dkfiCBYQroWCMq5EVrP6xqhwpump

[Vanta @VantaGames]

One Vanta match scores four things:
- how many turns it took
- damage dealt vs damage taken
- progress on the objective
- units kept alive

⚠️disclaimer: we will never dm you. Be aware of fake cas

[aquaz @0xaquaz]

This has been looping in my head all week: static benchmark starts decaying the moment it ships

Awnser key is training, and within a model or two you're measuring memory, not intelligence.

Vanta.

[Vanta @VantaGames]

@0xaquaz ⚔️⚔️

[OpenAI @OpenAI]

We’re sharing new research on a method for anticipating how models may behave in real-world use before release: simulating deployment with recent, de-identified user requests and studying candidate model responses. https://t.co/7RJzBfNniQ

[Vanta @VantaGames]

Benchmarks are saturating.

The gap between the top models is now smaller than the noise in the test itself.

Adversarial game environments are the way out of that.
It's what we're building.

[OpenAI @OpenAI]

Let’s talk about evals.

We’re always looking for better ways to measure and forecast model progress, especially as benchmarks get saturated or gamed.

@tejalpatwardhan, who leads our frontier evals team, spoke to @andrewmayne about why evals matter and what models need to be judged on next.