Arena preview (demo)

Synthetic visualisations that hint at the future arena experience with pairwise win rates and rating confidence bands.

Match-up win rate heatmap

Synthetic data showing the probability that model A wins a head-to-head match against model B (draws removed).

Model A / Model B	claude-sonnet-4-5-20250929-thinking-32k	claude-sonnet-4-5-20250929	gpt-5-high	gpt-5-chat	claude-opus-4-1-20250805-thinking-16k	chatgpt-4o-latest-20250326	gpt-4.5-preview-2025-02-27	gemini-2.5-pro	claude-opus-4-1-20250805	kimi-k2-0711-preview	deepseek-r1-0528	o3-2025-04-16	qwen3-max-preview	glm-4.6	qwen3-max-2025-09-23	qwen3-235b-a22b-instruct-2507
claude-sonnet-4-5-20250929-thinking-32k	-	0.66	0.74	0.64	0.74	0.56	0.60	0.66	0.61	0.49	0.62	0.54	0.62	0.43	0.48	0.54
claude-sonnet-4-5-20250929	0.66	-	0.70	0.60	0.60	0.61	0.65	0.58	0.56	0.52	0.64	0.51	0.48	0.55	0.51	0.41
gpt-5-high	0.62	0.75	-	0.66	0.77	0.67	0.72	0.65	0.58	0.69	0.67	0.54	0.55	0.55	0.49	0.47
gpt-5-chat	0.60	0.65	0.64	-	0.74	0.61	0.65	0.70	0.67	0.62	0.57	0.56	0.67	0.56	0.61	0.45
claude-opus-4-1-20250805-thinking-16k	0.59	0.75	0.69	0.62	-	0.74	0.78	0.77	0.68	0.67	0.66	0.67	0.53	0.59	0.66	0.52
chatgpt-4o-latest-20250326	0.68	0.62	0.65	0.79	0.77	-	0.69	0.72	0.67	0.68	0.57	0.66	0.67	0.66	0.66	0.50
gpt-4.5-preview-2025-02-27	0.68	0.62	0.58	0.71	0.76	0.64	-	0.70	0.73	0.65	0.60	0.57	0.70	0.61	0.64	0.57
gemini-2.5-pro	0.63	0.66	0.61	0.75	0.73	0.68	0.75	-	0.61	0.67	0.77	0.64	0.60	0.58	0.66	0.62
claude-opus-4-1-20250805	0.63	0.56	0.65	0.58	0.73	0.58	0.65	0.79	-	0.68	0.60	0.70	0.65	0.61	0.67	0.56
kimi-k2-0711-preview	0.59	0.58	0.68	0.61	0.61	0.60	0.67	0.68	0.68	-	0.77	0.62	0.66	0.67	0.56	0.65
deepseek-r1-0528	0.58	0.63	0.57	0.70	0.61	0.65	0.73	0.65	0.74	0.79	-	0.79	0.74	0.73	0.61	0.70
o3-2025-04-16	0.48	0.56	0.66	0.59	0.52	0.57	0.65	0.62	0.65	0.73	0.75	-	0.67	0.75	0.59	0.61
qwen3-max-preview	0.54	0.59	0.60	0.53	0.57	0.61	0.68	0.65	0.67	0.60	0.72	0.69	-	0.64	0.68	0.67
glm-4.6	0.47	0.55	0.56	0.50	0.59	0.57	0.64	0.72	0.56	0.72	0.74	0.79	0.73	-	0.63	0.76
qwen3-max-2025-09-23	0.58	0.62	0.47	0.61	0.56	0.64	0.51	0.65	0.68	0.60	0.62	0.59	0.59	0.67	-	0.75
qwen3-235b-a22b-instruct-2507	0.42	0.57	0.55	0.48	0.59	0.59	0.66	0.55	0.70	0.59	0.60	0.66	0.73	0.59	0.72	-

Roadmap notes

The production version will ingest real battle data, allow custom opponent selection and surface trend analytics to help researchers compare models with confidence.