Free GPU Inference

Run TRIBE v2
on a Free T4 GPU

Meta's fMRI foundation model, zero cloud cost. Google Colab's free T4 GPU delivers full TRIBE v2 inference — same accuracy as our RunPod production tier.

Request notebook access →
Specs
GPU
NVIDIA T4 (16GB VRAM)
Cost
$0 — 30h/week free quota
Download Time
3-5 min (model weights ~12GB)
Inference Time
~30s per text analysis

How It Works

1
Get Your HuggingFace Token
TRIBE v2 is a gated model on HuggingFace. Generate a free token — no payment method required. Accept the model terms at facebook/tribev2 once, then you're set.
2
Open the Notebook in Colab
Click "Open in Colab" above. The notebook connects to a free T4 GPU runtime automatically. (Runtime → Change runtime type → T4 GPU)
3
Run Cells 1-3 (One-Time Setup)
Paste your HF token, install dependencies, and download the model. Takes 3-5 minutes. The model stays cached in the Colab runtime.
4
Paste Content & Run Cells 4-7
Paste your brief, hook, or ad copy into TEXT_TO_ANALYZE. The notebook outputs full VAN/DMN/DAN/Limbic Z-scores with percentiles and downloads a JSON file.
5
Upload Results Back
On the Neormative main page, click "Upload Colab Results" and select the tribev2_results.json file. Your scores are applied immediately — same as RunPod production tier but at zero cost.
Why Colab vs RunPod
Colab (Free)
  • $0 / month
  • 30h/week GPU quota
  • Manual workflow
  • Best for development & testing
RunPod (Production)
  • ~$0.29/hr RTX 4090
  • Always-on API endpoint
  • Auto-scaling serverless
  • Best for client-facing use
Heuristic (Instant)
  • $0 / month
  • Instant (no GPU needed)
  • ~85% accuracy vs full TRIBE
  • Always available