AI leaderboard LLM rankings from the last 24 hours

19 models share first place — joint 1st of the 25 models on the Combined board in the last 24 hours — from Claude Sonnet 5.5 (95 ± 2.3) to GPT-6.1 Terra (89 ± 2.3), none of them measurably behind the leader. As of 30 Oct 2026, 15:14 UTC. The board is the 24-hour view: every model's runs over the last day, ranked with statistical ties. The best model for one job (coding, reasoning, agents, value or speed) is on best AI model.

Every model gets the same tasks on the same schedule (coding every four hours, reasoning and tool use daily, with an hourly canary probe) and is graded by running what it produces, not by votes or vendor-reported numbers. The Combined score is half coding, a quarter reasoning and a quarter tool use; a suite that is missing drops out and the others are reweighted. The methodology is public — every weight, statistical test and drift constant. The test tasks are private, so no lab can train on them. Every weight is on the methodology page.

Coding is re-tested every four hours, reasoning and tool use daily, and an hourly canary probe watches for sudden drops between runs. The 24-hour view averages each model's runs over the last day, so one odd attempt cannot move a place.

Every score carries a standard error from its own run-to-run repeatability. Walking down a board, a model joins the group above it unless that group’s leader is ahead by more than 1.96 times their combined standard error, so places such as 1, 1, 1, 4 are normal and a shared place is written joint. A model graded on fewer tasks than the rest keeps its position and is never called tied.

On the full leaderboard, with a free account: every model on every board over 24 hours, 7 days and 30 days, and a 3-model watchlist with email alerts. No card. Pro and up adds the newest run as it lands.

Claude Opus 5.5 and Claude Sonnet 5.5 share first place: joint 1st of the 25 models on the Coding board in the last 24 hours, each within the measurement noise of the leader (Claude Opus 5.5 97 ± 0.6, Claude Sonnet 5.5 95 ± 1.1). The Coding board is real debugging work graded by running each project's own tests; the method and the five best-placed models are on best LLM for coding.