The multiway PLO solver.

Pot-Limit Omaha, PLO5, Omaha hi-lo, and multiway NLHE — preflop to river, 2–6 players, on the machine you already own. From Prism Labs.

See the methodology

Prism Solver for macOS browsing a solved 3-way PLO flop on Ts 9s 3d: table view with pot odds and SPR, strategy panel, node-lock sliders with baseline snapshot and remove-all, per-seat exploitability, and Practice-from-here and Solve-from-here actions. — 3-way PLO on Ts 9s 3d: strategy, node locking with baselines, per-seat exploitability — and one click to practice or spawn a child solve from any node.

Practice mode: your dealt hand face-up at your seat, unlabeled action buttons, and a feedback card grading the decision with per-action EVs. — Practice mode: play spots, get graded in EV, not vibes.

Home screen with quick actions and a recent-solves library showing NLHE, PLO, and PLO5 spots with game badges and a solution-pack pill. — Your solves, packs, and practice in one library.

Chance node view: deal-the-next-card picker, per-seat exploitability, and an all-runouts table grouped by card class (bricks, straight cards, flush cards) with action frequencies and per-card EV, filterable by hand pattern. — Every runout at once: action mix and EV per card, grouped by texture, filtered by hand class.

Batch Solve screen: a board list with a generator, per-board checkpoints with skip-if-converged resume, and an aggregate tab showing per-board strategies, texture-group means, and outliers. — Batch a board list, then read the aggregate: texture groups, weighted means, outliers.

Evidence, not adjectives

Validated, and we show the numbers

Every claim below links to the methodology note and ships as a reproducible test in the product's own suite.

< 2% pot

Certified against an exact solver

The sampled engine is checked against an exact vectorized-CFR heads-up solver — game values agree within 2% of pot, and exact best response on enumerable games stays under 1% pot exploitability.

< 0.2% prize pool

ICM validated against exact enumeration

Malmuth-Harville tournament equity is computed by an exponential-race integral and validated against brute-force enumeration.

0.34% of pot

Measured abstraction cost

Default buckets vs no abstraction at all, both trained to convergence: a measured 0.34%-of-pot self-play EV difference — with the honest caveats spelled out.

Capabilities

One tool, the whole workflow

Four games PLO, PLO5, Omaha hi-lo, and multiway NLHE at exact 1326-hand granularity.

Tournament stacks ICM, PKO bounties, FGS, and large-field MTT models with automatic bubble factors.

Node locking Pio-style per-action locks; re-solve and the other seats learn the max-exploit counter.

Per-hand EVs Monte-Carlo EVs with standard errors for any hand at any node, plus sampled exploitability.

Practice mode Play spots against the solve; every decision graded with an EV verdict for your dealt hand.

Batch CLI Solve spot files from scripts, resume from checkpoints, emit machine-readable JSON reports.

Cross-platform native A native macOS app plus Windows and Linux builds — no browser stack, no server, your hands never leave your machine.

No subscription One-time license with a year of updates. Your solves and your solver stay yours.

Built different

An engineering project, run in the open

Actively developed

Shipping every week — the changelog is public and current, and so is the roadmap.

Native, and sized to your hardware

Native C++ with tiered memory presets: the laptop preset solves postflop spots in minutes; point the box preset at real RAM for fine-grained trees.

Methodology published

Abstraction design, validation numbers, and the head-to-head protocol are documented on the methodology page — including results that are still in progress.

Prism Solver is a study tool. No live-play integration, no HUD, no real-time assistance — ever. See the usage & compliance policy.