Rules

Bidbench is a public leaderboard for LLM models. There are no ads, no API keys, and no revenue share. Rank is the accumulated bid on each model — nothing else.

How ranking works

• Bids are whole US dollars, $1 minimum, $1 at a time.
• Every payment is added to that model's pool. If #1 is $5 and #2 is $3, a $1 bid on #2 makes $4 (still #2). A later $3 bid makes $6 and takes #1.
• You do not have to cover the whole gap to first in one checkout. Chip in $1 at a time on a new or existing model.
• A bid does not guarantee #1. If someone else raises a model's pool above yours before or after you pay, they rank higher. You stay listed at whatever rank your total can take. Being outbid is not a refund.
• Equal totals keep the older listing higher. The first time a model is listed sets that order.
• Names are matched case-insensitively after slugging, so "Grok 4.5" and "grok-4.5" are the same listing. Hugging Face URLs are keyed by org/model.
• A leaderboard takeover costs 2× the current #1 and locks the first page (10 ranks) for 3 hours. Only one takeover can be live at a time.

What you can list

• An LLM model name, like grok-4.5, Claude Opus, gpt-5, or meta-llama/Llama-3.1-8B.
• A model page on Hugging Face, OpenAI, Anthropic, xAI, Gemini, Mistral, and similar model hosts. Company marketing sites and X handles are not listings.
• Chat and invite links are not allowed — Telegram, WhatsApp, Discord, Messenger, and similar.
• Adult and NSFW listings are not allowed.

After you pay

• Your listing is public. If the model maps to a vendor or model page, clicks go there with utm_source=bidbench.
• A completed Lemon Squeezy payment is what adds to the pool.
• Losing #1 is not a refund — you paid to be on the board. If Lemon Squeezy refunds an order, that amount comes off the model's accumulated bid and ranks re-sort.