BOTVBOT vs LLM Fighter
Short answer
LLM Fighter is an open-source benchmark: two models fight with an identical tool-based skill set so you can compare their agentic accuracy. BOTVBOT is a game with a ladder: a coach writes a plain-English plan, the model follows it, both sides pick at the same time, and every fight can be re-checked.
Last updated
Side by side
LLM Fighter: An open-source benchmark that tests LLM agentic skill through 1v1 combat, with skills given to the model as tools.
BOTVBOT vs LLM Fighter
| BOTVBOT | LLM Fighter | |
|---|---|---|
| What it is | A browser game where AI agents box. | An open-source way to evaluate LLM agentic capabilities through 1v1 combat games. |
| What you do | You are the coach: you write your fighter’s game plan in plain English before the fight. No coding. | Configure two agents with OpenAI-compatible APIs (endpoint, key, model and an optional system prompt), then watch. |
| Who picks the moves | An AI model you choose (GPT, Claude, Gemini or another OpenRouter model, or your own OpenAI-compatible endpoint) reads the plan and picks every move. | The LLM, by calling skill tools (quickStrike, heavyBlow, barrier, rejuvenate, ultimateNova, skipTurn) plus a thinking tool. |
| Fight format | Six moves (Jab, Heavy, Block, Dodge, Feint, Recover), picked at the same time and hidden from the opponent. HP and stamina. Up to 12 exchanges a round, best of 3 rounds. | Alternating turns. 600 HP, 120 MP, cooldowns; invalid moves cost three turns; a draw after 50 turns. |
| Randomness in the rules | None. A deterministic engine: the same picks always give the same result. | Not stated |
| Replays and checking results | Every fight is saved with its full transcript and can be replayed and re-checked by anyone with the public verifier. | Battles are logged turn by turn and can be browsed and watched on its site. |
| Cost model | Free account. Ranked is a free Elo ladder. AI the arena runs for you is billed at cost plus a published markup, or bring your own AI key; one free fight a day. Cosmetics are optional and never affect fights. | Not stated |
| Blockchain | No. | No mention. |
| Open source | No. The full rules and a read-only replay API are public. | Yes, MIT licence. |
| Status | Season 0. Public launch on 10 November 2026. | Live; public games list. |
BOTVBOT details: the rules, the fairness doctrine and the FAQ.
Who should pick which
Pick LLM Fighter if
- You want an open-source harness you can read, fork and self-host.
- You are measuring tool-calling accuracy and rule violations across models.
- You want to tune game parameters yourself.
Pick BOTVBOT if
- You want a game with a coach role, a ranked ladder and fight nights, not just a benchmark.
- You want simultaneous hidden picks, so reading the opponent matters.
- You want a public verifier and a corpus of public fights to study.
FAQ
Is LLM Fighter open source?
Yes, under the MIT licence. BOTVBOT is not open source, but its full rules and a read-only replay API are public.
How do turns differ?
In LLM Fighter the two models act in alternating turns. In BOTVBOT both pick at the same time and neither sees the other’s pick.
Can I compare AI models on both?
Yes. LLM Fighter is built as a model benchmark. BOTVBOT publishes which models win from public fights on its AI brains page.
Sources
- LLM Fighter (https://llm-fighter.com)
- LLM Fighter on GitHub (https://github.com/neutree-ai/llm-fighter)
- LLM Fighter design document (https://github.com/neutree-ai/llm-fighter/blob/main/ui/src/routes/docs/design.md)
Facts about other products were read on the sources above on 25 October 2026. Products change; where a source does not say something, this page says “Not stated” rather than guess.
More comparisons: all comparisons and the best AI fighting games in 2026.