DeepSeek V4 Flash 0731: ARC-AGI Performance Analysis

Executive Summary: Released on July 31, 2026, the DeepSeek V4 Flash 0731 model demonstrates significant capabilities in abstract reasoning. When operating at peak computational effort, the model achieves a score of 89.0% on the ARC-AGI-1 Semi-Private benchmark and 61.4% on the ARC-AGI-2 Semi-Private benchmark.

🛠️ Model Specifications & Cost Efficiency

The model is available in three distinct reasoning configurations, allowing users to balance accuracy against computational cost.

Reasoning Variants: Max, High, and Low.
Resource Allocation:
- ARC-AGI-1 Semi-Private: $\text{Cost} = \$ 0.02 \text{ per task}$
- ARC-AGI-2 Semi-Private: $\text{Cost} = \$ 0.04 \text{ per task}$

Performance Scaling

The following diagram illustrates the correlation between reasoning effort and the resulting accuracy:

📊 Verified Benchmark Results

The table below details the verified scores across the different reasoning tiers:

Reasoning Variant	ARC-AGI-1 Score	ARC-AGI-2 Score	ARC-AGI-3 Score
Max	89.0%	61.4%	—
High	87.0%	56.0%	—
Low	84.0%	46.0%	—

~~Low Effort~~ $\rightarrow$ Max Effort results in a substantial jump in ARC-AGI-2 performance (from 46.0% to 61.4%).

🔍 Granular Task Evaluation

The following sections provide a task-by-task breakdown of the model's success ( $\checkmark$ ) or failure ( $\times$ ) across the three reasoning levels.

ARC-AGI-2 Public Evaluation (120 Tasks)

Task ID	Max Effort	High Effort	Low Effort
`0934a4d8`	$\times$	$\checkmark$	$\times$
`135a2760`	$\checkmark$	$\checkmark$	$\checkmark$
`136b0064`	$\checkmark$	$\checkmark$	$\checkmark$
`13e47133`	$\times$	$\times$	$\times$
`142ca369`	$\times$	$\checkmark$	$\times$
`16b78196`	$\times$	$\times$	$\times$
`16de56c4`	$\checkmark$	$\checkmark$	$\checkmark$
`1818057f`	$\checkmark$	$\checkmark$	$\checkmark$
`195c6913`	$\checkmark$	$\times$	$\times$
`1ae2feb7`	$\checkmark$	$\checkmark$	$\checkmark$
`20270e3b`	$\checkmark$	$\checkmark$	$\times$
`20a9e565`	$\times$	$\times$	$\times$
`21897d95`	$\times$	$\times$	$\times$
`221dfab4`	$\checkmark$	$\checkmark$	$\times$
`247ef758`	$\checkmark$	$\checkmark$	$\checkmark$
`269e22fb`	$\times$	$\times$	$\times$
`271d71e2`	$\times$	$\checkmark$	$\times$
`28a6681f`	$\checkmark$	$\times$	$\times$
`291dc1e1`	$\checkmark$	$\checkmark$	$\times$
`2b83f449`	$\times$	$\times$	$\times$
`2ba387bc`	$\checkmark$	$\checkmark$	$\checkmark$
`2c181942`	$\checkmark$	$\checkmark$	$\times$
`2d0172a1`	$\times$	$\times$	$\times$
`31f7f899`	$\checkmark$	$\checkmark$	$\checkmark$
`332f06d7`	$\checkmark$	$\times$	$\times$
`35ab12c3`	$\checkmark$	$\checkmark$	$\checkmark$
`36a08778`	$\checkmark$	$\checkmark$	$\checkmark$
`38007db0`	$\checkmark$	$\checkmark$	$\checkmark$
`3a25b0d8`	$\times$	$\times$	$\times$
`3dc255db`	$\times$	$\times$	$\times$
`3e6067c3`	$\checkmark$	$\times$	$\times$
`409aa875`	$\checkmark$	$\checkmark$	$\times$
`446ef5d2`	$\times$	$\times$	$\times$
`45a5af55`	$\checkmark$	$\checkmark$	$\checkmark$
`4a21e3da`	$\checkmark$	$\times$	$\checkmark$
`4c3d4a41`	$\times$	$\checkmark$	$\checkmark$
`4c416de3`	$\times$	$\times$	$\times$
`4c7dc4dd`	$\times$	$\times$	$\times$
`4e34c42c`	$\times$	$\times$	$\times$
`53fb4810`	$\checkmark$	$\checkmark$	$\checkmark$
`5545f144`	$\times$	$\times$	$\times$
`581f7754`	$\times$	$\times$	$\times$
`58490d8a`	$\checkmark$	$\checkmark$	$\checkmark$
`58f5dbd5`	$\checkmark$	$\checkmark$	$\checkmark$
`5961cc34`	$\checkmark$	$\checkmark$	$\checkmark$
`5dbc8537`	$\checkmark$	$\times$	$\times$
`62593bfd`	$\times$	$\times$	$\times$
`64efde09`	$\times$	$\checkmark$	$\times$
`65b59efc`	$\checkmark$	$\checkmark$	$\checkmark$
`67e490f4`	$\checkmark$	$\checkmark$	$\checkmark$
`6e453dd6`	$\checkmark$	$\checkmark$	$\checkmark$
`6e4f6532`	$\checkmark$	$\times$	$\times$
`6ffbe589`	$\times$	$\times$	$\times$
`71e489b6`	$\checkmark$	$\checkmark$	$\times$
`7491f3cf`	$\checkmark$	$\times$	$\checkmark$
`7666fa5d`	$\checkmark$	$\checkmark$	$\times$
`78332cb0`	$\times$	$\times$	$\times$
`7b0280bc`	$\checkmark$	$\times$	$\times$
`7b3084d4`	$\checkmark$	$\checkmark$	$\times$
`7b5033c1`	$\times$	$\times$	$\checkmark$
`7b80bb43`	$\times$	$\times$	$\times$
`7c66cb00`	$\times$	$\times$	$\checkmark$
`7ed72f31`	$\checkmark$	$\checkmark$	$\times$
`800d221b`	$\times$	$\times$	$\times$
`80a900e0`	$\checkmark$	$\checkmark$	$\checkmark$
`8698868d`	$\times$	$\times$	$\times$
`88bcf3b4`	$\times$	$\times$	$\times$
`88e364bc`	$\times$	$\times$	$\times$
`89565ca0`	$\checkmark$	$\times$	$\checkmark$
`898e7135`	$\checkmark$	$\checkmark$	$\checkmark$
`8b7bacbf`	$\times$	$\times$	$\times$
`8b9c3697`	$\checkmark$	$\times$	$\times$
`8e5c0c38`	$\checkmark$	$\times$	$\checkmark$
`8f215267`	$\checkmark$	$\checkmark$	$\checkmark$
`8f3a5a89`	$\checkmark$	$\checkmark$	$\checkmark$
`9385bd28`	$\checkmark$	$\checkmark$	$\times$
`97d7923e`	$\checkmark$	$\times$	$\checkmark$
`981571dc`	$\checkmark$	$\checkmark$	$\checkmark$
`9aaea919`	$\checkmark$	$\checkmark$	$\checkmark$
`9bbf930d`	$\times$	$\times$	$\times$
`a251c730`	$\checkmark$	$\checkmark$	$\checkmark$
`a25697e4`	$\times$	$\times$	$\times$
`a32d8b75`	$\times$	$\times$	$\times$
`a395ee82`	$\checkmark$	$\checkmark$	$\checkmark$
`a47bf94d`	$\times$	$\checkmark$	$\checkmark$
`a6f40cea`	$\checkmark$	$\checkmark$	$\checkmark$
`aa4ec2a5`	$\checkmark$	$\checkmark$	$\checkmark$
`abc82100`	$\times$	$\times$	$\times$
`b0039139`	$\checkmark$	$\times$	$\checkmark$
`b10624e5`	$\checkmark$	$\checkmark$	$\checkmark$
`b5ca7ac4`	$\checkmark$	$\checkmark$	$\checkmark$
`b6f77b65`	$\checkmark$	$\checkmark$	$\times$
`b99e7126`	$\checkmark$	$\checkmark$	$\checkmark$
`b9e38dc0`	$\times$	$\times$	$\times$
`bf45cf4b`	$\