The AI Race
Watch the AI race unfold
Scrub the timeline to travel through 18 months of model releases. See who held the crown, who climbed the ranks, and how the benchmark landscape evolved.
May 20263 releases
Monthly Race Snapshot
Export or embed the live race view for the currently selected month.
Crown Holder
No ranked release this month
Runner Up
Only one release this month
Releases this month
3 modelsProvider Race
Cumulative avg. top-3 score through May 2026Benchmark Health
How fresh are the benchmarks we use to score models? Green means the benchmark is actively separating models. Red means scores are bunching up or the benchmark is outdated.
Agentic
35 benchmarks32 current3 refreshing
Coding
22 benchmarks18 current3 refreshing1 stale1 saturated
Reasoning
22 benchmarks18 current1 refreshing3 stale1 saturated
Multimodal
45 benchmarks40 current5 refreshing
Knowledge
27 benchmarks20 current3 refreshing4 stale1 saturated
Multilingual
8 benchmarks7 current1 stale
Instruction Following
3 benchmarks2 current1 stale
Math
23 benchmarks17 current3 refreshing3 stale
Want deeper historical analysis?
Explore 21 months of Arena Elo ratings — crown changes, provider dominance, open-source gap tracking, and more.
LLM Leaderboard History