Get the weekly radar
One short email with the freshest products, every week.
Foresight outputs calibrated probabilities instead of plausible text, built for forecasting.
Grounded in available product and source data
An AI forecasting model for calibrated predictions is a narrower, more specific pitch than a general chat model, and that's the entire point of Foresight from Lightning Rod Labs: it's built to output calibrated probabilities rather than plausible-sounding text, positioned as an alternative to frontier models specifically for prediction tasks.
The training method behind that distinction is named directly — 'Future-as-Label,' presented at the ICML 2026 AI Forecasting Workshop — and the model pairs it with an automatic research capability that gathers relevant context before producing an estimate, rather than forecasting from a static training cutoff alone. Named use cases lean commercial and operational: prediction-market automation, market-maker applications, risk forecasting from news and regulatory filings, and event monitoring with live probability updates as new information arrives.
The benchmark numbers are specific rather than vague marketing claims: measured by Brier Skill Score against resolved Polymarket questions, Foresight v4 (Full) scores 26%, Foresight v4 (Low) scores 13%, and Foresight v3 scores 19%, with the site stating it outperforms GPT-5.4, GPT-5, Opus 4.6, and Gemini 3.1 Pro on this specific measure. Cost efficiency gets the same treatment — GPT-5 is stated to cost 1.7 times more per output token, and Gemini 3.1 Pro 2.0 times more, with an all-in cost advantage demonstrated across a batch of 1,000 forecasts.
Access runs through an OpenAI-compatible API using the model name 'foresight-v4,' meaning existing OpenAI client code can often point at Foresight with minimal changes, plus extensions like a research flag for the auto-context-gathering feature. What's worth flagging directly: there's no published freemium or flat-subscription tier shown on the pages reviewed — pricing is usage-based per output token, so cost for a specific workload depends on forecast volume and would need confirming through a demo or the dashboard rather than a fixed plan.
Foresight is trained specifically to output calibrated probabilities using a method the site calls 'Future-as-Label,' rather than being a general-purpose chat model adapted to forecasting after the fact — and the site backs that distinction with Brier Skill Score benchmarks against GPT-5.4, GPT-5, Opus 4.6, and Gemini 3.1 Pro.
It can research beyond its training data — an automatic research capability auto-gathers relevant context before producing a probability estimate, which matters for forecasting questions tied to fast-moving news or filings.
There's no published no-cost tier to fall back on — pricing runs on a usage basis per output token instead of a flat plan, so testing it at scale means confirming likely cost directly through the dashboard or a booked demo first.
In many cases yes — Foresight exposes an OpenAI-compatible API interface using the model name 'foresight-v4,' so existing OpenAI client setups can often point at it with minimal changes.
Enterprise, government, and startup clients working on prediction-market automation, market-making, risk forecasting, or quant signal tracking, per the named use cases and client logos shown on the site, rather than casual or individual forecasting use.
These 1 comments were collected from the discovery source. ProductVisit does not host replies in Release A.
Hey Product Hunt — Ben here, founder of Lightning Rod Labs. Frontier AI is powerful, but it is not built for forecasting. Frontier models are trained to produce plausible text, not well-calibrated probabilities about what will actually happen. They are also expensive to run inside agentic workflows, where bots may need to forecast thousands of markets, events, or decisions. We trained Foresight to make better predictions at lower inference cost. Foresight is an AI forecasting API with better accuracy at a lower inference cost. It is trained using our Future-as-Label method (Spotlight at the ICML 2026 AI Forecasting Workshop), which uses real-world outcomes over time for training. Instead of hand-labeling datasets or imitating generic text, Foresight learns from what actually happened. Foresight beats frontier models 100x larger on live prediction benchmarks, like ProphetArena and ForecastBench, with a particularly large lead in prediction market categories like Sports & Politics. Our API is OpenAI-compatible, so developers can easily swap it into existing workflows. Better accuracy. Cheaper inference. OpenAI-compatible API. Use code PHFORESIGHT for $50 free API credits this month. We'd love feedback from builders working on forecasting agents, prediction tools, or any workflow where better forecasts matter.
3 source upvotes