Predict where any token goes —
verified on-chain.
Select an OpenGradient model, pick any Binance token, get a 4-hour forecast with stop-loss and take-profit. Every inference signed by your wallet, settled in OPG, and recorded on Base Sepolia.
Choose a model
Every model card links to its real page on hub.opengradient.ai
OG 1hr Volatility
Predicts standard deviation of 1-minute returns over the next hour. Originally trained on ETH/USDT Jul 2023–Aug 2024 with ρ=0.85 out-of-sample correlation. Ideal for volatility sizing, options pricing, and dynamic SL/TP.
Allora 5min Volatility
Random Forest model predicting ETH/USD 5-minute forward volatility from 10 minutes of OHLC data. Built for high-frequency entry timing and intraday risk control.
Vanna Volatility Risk
Two-stage model: K-means clusters tokens by historical volatility profile (1h/12h/24h, 5m granularity), then XGBoost computes a risk z-score relative to its class. Flags outliers fast.
Pump & Dump Detector
Verifiable binary classifier. Flags tokens showing pump-and-dump manipulation patterns via 5 normalised features. Score ≥0.5 = manipulation signal. Every inference recorded on-chain.
PolyEdge Analyzer
AI-powered Polymarket prediction market analyzer. Computes risk scores (liquidity, volatility, time, resolution) and estimates true probability vs market price to find edge. Only meaningful when a matching PM market exists.
OG ETH 1h Forecast
TCN (Temporal Convolutional Network) that emits the predicted % return for ETH over the next hour. Ideal for Polymarket up-or-down markets with 1h resolution.
OG BTC 1h Forecast
TCN architecture trained on BTC/USDT hourly OHLCV. Emits predicted % return over next hour — positive = UP, negative = DOWN. Direct mapping to Polymarket up/down hourly markets.
OG ETH 1m Return
Log-price Lasso model for ETH/USDT 1-minute return. OHLC features are log-transformed, passed through a single-layer weighted model, then post-processed via arctanh + inverse-Z. Positive output = long, negative = short. Trained on unshuffled time-ordered splits with long/short cumulative-return objective.
OG ETH 5m Return
Log-price Lasso model for ETH/USDT 5-minute return. Takes the 10 most recent 5-minute OHLC candles and emits a signed return forecast. Same architecture as the 1m and 15m variants — positive = long, negative = short.
OG ETH 15m Return
Log-price Lasso model for ETH/USDT 15-minute return. Takes the 10 most recent 15-minute OHLC candles and emits a signed return forecast. Same family as the 1m and 5m variants — positive = long, negative = short.
ETH 30m Lasso Forecast
Lasso regression optimized for cumulative returns of a long/short strategy (positive forecast = long, negative = short). Takes 10 recent 30-minute ETH/USDT candles and emits the forecast 30-min return.