Live Evals Running

Crowdfund AI benchmarks.
Watch them run live.

Community-funded evaluations run as far as contributions allow. Partial results shown in real-time. Pause when depleted. Resume when funded.

Snitch Bench
GPT-4 Turbo
Running
Funding Progress $94 / $130
847 samples
32 contributors
Simple Bench
Claude Sonnet 4.5
Running
Funding Progress $412 / $850
1,243 samples
89 contributors
Long Context
Gemini 2.0 Flash
Running
Funding Progress $1,847 / $2,000
2,891 samples
156 contributors

How It Works

01

Choose a benchmark

Browse active and requested evaluations. Each benchmark shows current funding progress and partial results.

02

Contribute any amount

Funds go to escrow and are released as API requests complete. Evals run as far as funding allows, pause when depleted, resume when funded.

03

Watch results live

Partial results display in real-time. All results are public. Raw data is published for download.

For Maintainers

Your benchmark. Funded and running.

Submit your eval and earn 5% commission on every contribution
Platform handles execution and payments - you focus on the benchmark
Run against frontier models every release (~15 per year)
Submit Your Benchmark
For Sponsors

Fund the evals that matter to your stack

Fund specific benchmarks or models at scale
Support transparent AI evaluation
Become a Sponsor
For Researchers

Get your eval in front of the models

Request benchmarks be added - community signals interest through funding
Your research runs against frontier models (~15 releases per year)
Request a Benchmark

Why This Matters

Problem

Running AI benchmarks costs hundreds to thousands of dollars. Snitch Bench costs $130 to run. Artificial Analysis runs cost $2,000. Researchers cannot re-run evaluations on every new model.

$2,000 Cost for a single Artificial Analysis Grok 4 run

Solution

Crowdfunding keeps evals current. For Snitch Bench, only 48 people at $3 each funds a complete benchmark run. That is 0.01% of Theo.gg's 475K subscribers.

48 people at $3 each to fund a $130 benchmark run