RLServ AI - Poker Intelligence Platform

Welcome to RLServ, a high-performance AI inference service for poker strategy. Our platform provides real-time decision-making support across multiple poker variants with state-of-the-art neural network models.

Throughput
6,500QPS
P95 Latency
35ms
Models
15+
Game Types
9

🎮 Texas Hold'em Models

Our most comprehensive suite of models, supporting various Hold'em formats with specialized neural networks for each variant.

Cash Game Universal
v3.2.1
Production
cash_general_formats_702-FIXEV.onnx
Universal cash game model supporting 2-9 players across all stake levels with observation version 702
P95 Latency
45ms
Throughput
6,500QPS
Default Profile: Game Theory Optimal
VPIP: 25%
PFR: 17%
Supported Games:
nlhe-cash
cash-game
6-max
full-ring
heads-up
nlhe
View Full Details & API Docs
MTT Tournament
v2.8.5
Production
mtt-normal-dp-702.onnx
Multi-table tournament model with ICM calculations for normal MTT, bounty, PKO, and mystery bounty formats
P95 Latency
62ms
Throughput
5,200QPS
Default Profile: GTO with ICM
VPIP: 22%
PFR: 15.5%
Supported Games:
nlhe-mtt
mtt
tournament
icm
pko
bounty
View Full Details & API Docs

🍍 Pineapple Poker Models

Three-card starting hand variants with specialized decision-making for discard optimization.

Classic Pineapple
v1.2.0
Beta
pineapple-classic-v1.onnx
3-card starting hand variant with pre-flop discard optimization
P95 Latency
38ms
Throughput
4,200QPS
Default Profile: GTO Pineapple
VPIP: 28.5%
PFR: 19.2%
Supported Games:
pineapple-classic
pineapple
classic
discard
3-card
View Full Details & API Docs

🀄 Guandan (掼蛋) Models

Partnership card game popular in China, featuring team-based strategy and bomb combinations.

🚀 Getting Started

API Endpoint

POST /api/strategy_act
Authorization: Bearer YOUR_TOKEN

Quick Example

{
  "request_id": "unique-id",
  "hand": {
    "game_type": "nlhe",
    "players": [...],
    "board": ["Ah", "Kc", "7d"]
  },
  "strategy_profile": {
    "target_profile_name": "gto"
  }
}

Support