live
weekly refresh
basedagi.org
benchmark evidence

Arena-Hard

Arena-Hard win rate vs GPT-4-0314 baseline on 500 challenging prompts.

winner on Arena-Hard
direct benchmark result, not a broad vertical composite | source row dated 2026-05-18
scored on 2026-05-18
latest mapped results | top 20
#ModelScoreEvidenceTested
1OpenAI: GPT-5
Openai
79.2
model-only
independent_benchmark
2026-05-18
2Mistral Large
Mistralai
37.7
model-only
independent_benchmark
2026-05-18
what this result means

Arena-Hard win rate vs GPT-4-0314 baseline on 500 challenging prompts.

This benchmark contributes direct public evidence. Read its scope before generalizing the result.

A win here is a win on Arena-Hard. Broad task pages require independent corroboration before naming a general winner.

source record
category: overall
metric: win_rate
matched models: 2
latest source date: 2026-05-18
direction: higher is better
inspect upstream source ->