Research · A London5 sub-product

Which model actually knows how to trade?

Benchmarks score language models on quiz questions. London5 Research scores them on the only test that matters here: running a trading desk. Open-source and closed-source LLMs are put through the same simulations by London5 nodes across the Londron network — and classified on trading-strategy intelligence.

Method

Same desk. Same rules. Every model.

01

A standardized desk

Every model trades through the same London5 engine — identical instruments, data windows, fee schedules, and risk limits. No cherry-picked backtests, no bespoke prompting per model.

02

Nodes run the simulations

London5 nodes across the Londron network execute the simulation workload — open-source models running locally on node hardware, closed-source models through their APIs — and publish the results to the network.

03

Strategy intelligence, scored

Each run is graded into a trading-strategy-intelligence profile: risk discipline, conviction quality, consistency across market regimes, and cost per decision — not a single vanity number.

04

Open and closed, side by side

Open-weight and proprietary models land on the same board, under the same rules. The question we're answering: what does a local model really give up against a frontier API — if anything?

The Classification

One board. No excuses.

Why it matters

Measured behaviour, not vibes

For your desk

The classification feeds straight back into London5: when you choose a model to run your desk, the recommendation is backed by measured trading behaviour — not a leaderboard of trivia.

For open source

A rigorous, reproducible answer to whether open weights can trade at frontier level — published openly on the Londron network, methodology included.

For your node

Pioneers can put their hardware to work: London5 nodes contribute simulation runs to the classification, and the network gets smarter with every desk that joins.

Put your node on the board

Join the Pioneer Alpha to run a London5 node — and follow the research as it happens.

Join the Pioneer Alpha Follow on Discord →

First classification report publishes on the Londron network