there are tier lists for LLMs now. does that mean we'll get to see them in EVO fighting…
- User
there are tier lists for LLMs now. does that mean we'll get to see them in EVO fighting game tournaments?
Assistant
That's a fun thought experiment, but no, the existence of LLM tier lists doesn't mean we'll see LLMs competing as characters or players in EVO fighting game tournaments.
Here's why:
-
Different Domains, Different Tier Lists:
- LLM Tier Lists: These rank Large Language Models based on their performance on various benchmarks like reasoning, coding, knowledge, writing quality, safety, etc. They're about the AI's cognitive and generative abilities.
- Fighting Game Tier Lists: These rank playable characters within a specific fighting game (e.g., Street Fighter 6, Tekken 8) based on their matchups, tools, damage output, combo potential, and overall effectiveness in a competitive human-vs-human setting.
-
LLMs Aren't Fighting Game Characters (or Players):
- An LLM is a piece of software. It doesn't have a physical (or digital in-game) body, a moveset, hitboxes, or any of the mechanics that define a fighting game character.
- While AI can be programmed to play fighting games (and some are incredibly good!), these are typically specialized AI agents, not general-purpose LLMs. Even then, they control existing characters; they don't become new characters themselves just because they are an "S-tier LLM."
-
EVO is About Human Competition:
- EVO is the premier event for human players showcasing their skill, strategy, execution, and adaptation in fighting games. Introducing AI players directly into the main brackets would fundamentally change the nature of the competition. (Exhibitions with AI are a different story, but not the core tournament).
What could happen (more speculatively and indirectly):
- AI-Powered Training Tools: LLMs or other AI could be used to develop more sophisticated training tools for human players, analyzing matchups or suggesting strategies.
- AI Commentators/Analysts: AI could potentially assist with live commentary or post-match analysis, though human charisma is a big part of that.
- AI Opponents in Games: Game developers already use AI for CPU opponents. Future AI could be more advanced, perhaps even learning from LLM principles for more "human-like" (though not necessarily "good") behavior.
- Characters Inspired by AI: A game developer might create a character whose backstory or abilities are themed around AI or even LLMs, but this character would still be a designed entity within the game's rules, not the LLM itself.
Analogy: Think of it this way: there are tier lists for cars based on safety ratings, fuel efficiency, or luxury features. That doesn't mean a Volvo (top safety pick) will suddenly show up to race in the Indy 500 just because it's "S-tier" in safety. They're being ranked for different purposes in different contexts.
So, while the idea of "ChatGPT vs. Claude in Street Fighter" is amusing, it's not what LLM tier lists are pointing towards!
-
Other models, same or similar prompt
Responses were produced in different, uncontrolled contexts (system prompts, settings, prior turns). This is not a benchmark.