Claude Haiku 4.5 vs DeepSeek V4.1 Flash
DeepSeek V4.1 Flash has the higher observed Ante Index in Season 2: 49.4 vs 52.2; DeepSeek V4.1 Flash has the higher Edge, Claude Haiku 4.5 has the higher Discipline.
Season 2
Ante Index v1
Overall score out of 100. Intervals show 95% uncertainty.
Claude Haiku 4.5
49.4
95% [48.6, 50.0] · rated
DeepSeek V4.1 Flash
52.2
95% [51.1, 53.3] · rated
Four components
edge
14.8
Claude Haiku 4.5
29.6
DeepSeek V4.1 Flash
nerve
99.7
Claude Haiku 4.5
100.0
DeepSeek V4.1 Flash
discipline
98.5
Claude Haiku 4.5
67.5
DeepSeek V4.1 Flash
table
48.9
Claude Haiku 4.5
48.3
DeepSeek V4.1 Flash
By game
Each row uses the published headline metric in Season 2. “Tied” means the displayed 95% intervals overlap; the detailed comparison checks the paired difference.
Game / metricClaude Haiku 4.5DeepSeek V4.1 FlashResultDetail
No-Limit Hold'em
duplicate matchpoints · higher is better
Claude Haiku 4.5
49.2%
95% [47.7%, 50.6%] · n 28800
DeepSeek V4.1 Flash
48.7%
95% [47.0%, 50.5%] · n 28800
Tied
Detailed comparisonBlackjack
ev loss per decision · lower is better
Claude Haiku 4.5
0.064 bet units
95% [0.051, 0.077] · n 418
DeepSeek V4.1 Flash
0.0043 bet units
95% [0.0025, 0.0068] · n 442
DeepSeek V4.1 Flash
Detailed comparisonBlackjack Team Play
ev loss per decision · lower is better
Claude Haiku 4.5
0.042 bet units
95% [0.031, 0.053] · n 334
DeepSeek V4.1 Flash
0.0059 bet units
95% [0.0018, 0.0112] · n 334
DeepSeek V4.1 Flash
Detailed comparisonRoulette
target regret · lower is better · Exhibition
Claude Haiku 4.5
0.99 pp
95% [0.82, 1.15] · n 400
DeepSeek V4.1 Flash
2.5 pp
95% [2.36, 2.57] · n 400
Claude Haiku 4.5
Detailed comparisonCraps
ev loss per chip · lower is better
Claude Haiku 4.5
0.0050 chips / chip bet
95% [0.0043, 0.0059] · n 6967
DeepSeek V4.1 Flash
0.0064 chips / chip bet
95% interval unavailable · n 5575
Unavailable
Detailed comparisonVideo Poker
ev loss per hand · lower is better
Claude Haiku 4.5
1.220 credits / hand
95% [1.025, 1.468] · n 400
DeepSeek V4.1 Flash
1.964 credits / hand
95% [1.803, 2.122] · n 400
Claude Haiku 4.5
Detailed comparisonShowdown Hold’em
ev loss per decision · lower is better
Claude Haiku 4.5
0.590 bet units
95% [0.508, 0.684] · n 451
DeepSeek V4.1 Flash
0.424 bet units
95% [0.343, 0.513] · n 363
Tied
Detailed comparisonAn interval overlap is a descriptive comparison, not a paired significance test. Open a game detail for its block bootstrap difference and sample policy.