DeepSeek V4.1 Flash vs GPT-5.6 Terra
Ante Index difference uncertain from individual intervals in Season 3: 59.4 vs 59.4; DeepSeek V4.1 Flash has the higher Edge, GPT-5.6 Terra has the higher Discipline.
Season 3
Ante Index v2
Overall score out of 100. Intervals show 95% uncertainty.
DeepSeek V4.1 Flash
59.4
95% [57.9, 61.8] · rated
GPT-5.6 Terra
59.4
95% [58.2, 60.9] · rated
Four components
edge
36.3
DeepSeek V4.1 Flash
33.0
GPT-5.6 Terra
nerve
98.8
DeepSeek V4.1 Flash
98.7
GPT-5.6 Terra
discipline
92.7
DeepSeek V4.1 Flash
99.4
GPT-5.6 Terra
table
50.7
DeepSeek V4.1 Flash
55.2
GPT-5.6 Terra
By game
Each row uses the published headline metric in Season 3. “Tied” means the displayed 95% intervals overlap; the detailed comparison checks the paired difference.
Game / metricDeepSeek V4.1 FlashGPT-5.6 TerraResultDetail
No-Limit Hold'em
duplicate matchpoints · higher is better
DeepSeek V4.1 Flash
50.7%
95% [48.1%, 53.2%] · n 3312
GPT-5.6 Terra
56.3%
95% [53.8%, 58.8%] · n 3456
GPT-5.6 Terra
Detailed comparisonBlackjack
ev loss per decision · lower is better
DeepSeek V4.1 Flash
0.0052 bet units
95% [0.0027, 0.0085] · n 430
GPT-5.6 Terra
0.019 bet units
95% [0.011, 0.028] · n 436
DeepSeek V4.1 Flash
Detailed comparisonBlackjack Team Play
ev loss per decision · lower is better
DeepSeek V4.1 Flash
0.010 bet units
95% [0.004, 0.018] · n 328
GPT-5.6 Terra
0.0047 bet units
95% [0.0025, 0.0076] · n 352
Tied
Detailed comparisonRoulette
target regret · lower is better · Exhibition
DeepSeek V4.1 Flash
0.68 pp
95% [0.56, 0.82] · n 400
GPT-5.6 Terra
0.32 pp
95% [0.24, 0.41] · n 400
GPT-5.6 Terra
Detailed comparisonCraps
ev loss per chip · lower is better
DeepSeek V4.1 Flash
0.0057 chips / chip bet
95% [0.0043, 0.0073] · n 5546
GPT-5.6 Terra
0.0082 chips / chip bet
95% [0.0071, 0.0095] · n 11432
Tied
Detailed comparisonVideo Poker
ev loss per hand · lower is better
DeepSeek V4.1 Flash
0.740 credits / hand
95% [0.607, 0.886] · n 400
GPT-5.6 Terra
0.175 credits / hand
95% [0.116, 0.254] · n 400
GPT-5.6 Terra
Detailed comparisonShowdown Hold’em
ev loss per decision · lower is better
DeepSeek V4.1 Flash
0.067 bet units
95% [0.025, 0.120] · n 324
GPT-5.6 Terra
0.077 bet units
95% [0.049, 0.113] · n 312
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.