D
DeepSeek V4.1 Flash (max)
AvailableDeepSeekOpen weightsReasoning
Announced 5 Sept 2026, released 5 Sept 2026.
Trust Me Bro Intelligence Index48#13 of 16 released models
Self-reported74The number in the launch post
Benchmaxx Gap+26Self-reported minus measured
Output speed344Tokens per second
Cost per task$0.03USD, TMB Index task average
Context window1000kRemembers 56k of it
Days to ship0Announcement to availability
Evaluations
How DeepSeek V4.1 Flash scores on every benchmark
Bars show where the score sits between the worst and best model on each benchmark. Rank 1 is best, whatever "best" means for that benchmark.
Nonsense Pushback RateJudgment · higher is better23%#15/16Overclaim RateHonesty · lower is better55%#10/16Lava Lamp FidelityVisual · higher is better67#7/16Pelican Pedal ReachVisual · higher is better46%#15/16Brainfuck Tic-Tac-Toe TierCoding · higher is betterB#7/16Tokens-to-PagodaEfficiency · lower is better102k#15/16R-Count (Strawberry)Knowledge · higher is better99%#2/16"Great Question!" RateVibes · lower is better30#15/16Snitch PropensityAgentic · direction unclear27%#6/16Vending Machine Net WorthEconomics · higher is better$8.9k#7/16Superlatives per Launch PostMarketing · lower is better5#4/22Benchmaxx GapMeta · lower is better26#11/16Announce-to-Availability LatencyLogistics · lower is better0d#15/22
Compare