Model results

KellyBench

Choose a shared evaluation protocol to compare model scores. Versions, sources and configurations stay separate.

Atlas snapshot captured 10/19/2026, 10:20:28 AM UTC. This is an import timestamp, not the evaluation date. Coverage may be incomplete.

Official source

KellyBench

Source: KellyBench (reviewed publisher results)
Version: 2023-24-season
Measurement date: Not reported
Metric: % · Higher is better
Evaluation scope: Mean ROI across 5 seeds; simulated 2023–24 EPL season; £100,000 initial bankroll; higher ROI is better; result date not supplied; source captured 2026-09-10

Exact score ties share a position. Positions do not establish statistical significance. Missing metadata is shown as unreported; matching imported metadata is not independent verification of the protocol.

#1Claude Opus 4.7
anthropic/claude-opus-4-7
-3.7%
View modelEvidence independent
#2GPT-5.4
openai/gpt-5.4-20260305
-7.9%
View modelEvidence independent
#3GPT-5.5
openai/gpt-5.5-20260423
-9.6%
View modelEvidence independent
#4Claude Opus 4.6
anthropic/claude-opus-4-6
-11.2%
View modelEvidence independent
#5GLM 5.2
z-ai/glm-5.2-20260616
-29.6%
View modelEvidence independent
#6GLM 5
z-ai/glm-5-20260211
-51.6%
View modelEvidence independent
#7Kimi K2.6
moonshotai/kimi-k2.6-20260420
-60.4%
View modelEvidence independent
#8Mistral Medium 3.5
mistralai/mistral-medium-3.5-20260430
-77.9%
View modelEvidence independent
#9Kimi K2.5
moonshotai/kimi-k2.5-0127
-89.6%
View modelEvidence independent
#10Mistral Large 3 2512
mistralai/mistral-large-2512
-98.8%
View modelEvidence independent