1/ The era of community-powered AI rankings is here. Over 5 days, the Recall community crowdsourced an ungameable benchmark to test 50 popular AI models and cast 7.8M votes in the world's largest pre-launch skill prediction for GPT-5 on . Let's dig in🧵

2/ First, the Recall Predict stats are just massive. 🗓️ 5 days since launch 👀 132K user signups 🤖 50 AI models tested ⚡️ 21K skills + tests submitted 🔥 7.8M predictions made

3/ Crowdsourcing AI's ungameable benchmark Existing benchmarks are broken: ❌Models train on them ❌Misaligned to user needs ❌Black-box Users submitted 7K skills + 13.5K tests to create an ungameable benchmark that measures and ranks AI models on skills that matter to them.

4/ Predicting AI model performance Users made 7.8M predictions on how GPT-5 and 50 other models would perform against their benchmark. Top picks: 1. @OpenAI GPT-5 2. @Google Gemini 2.5 Pro 3. @xAI @Grok 4 Predictions are now closed. Visit for results.

5/ Now that GPT-5 has launched, what's next? Run the benchmark 📊 Step 1. Measure AI model performance Step 2. Publish results and leaderboard Step 3. Award points for contributions Build the next benchmark 👇

6/ Recall Predict is just getting started. Head to to shape the next ungameable, community-powered benchmark for upcoming AI models like @Google Gemini 3. AI development doesn't stop. And it should be shaped by the people.