readable AI benchmarks (simplified)

readable AI benchmarks

DeepSeek

DeepSeek V4 Pro (high)

reasoning · open weights · Apr 24, 2026

DeepSeek V4 Pro (high) has a Quality Score of 50.6, ranking 32nd among 53 scored models. Per 1M tokens, pricing is $0.435 input, $0.870 output, and $0.0036 cached input. Its Reliability Score is 45.1, ranking 35th among 53 scored models, below average (total average: 50.1). Its Value Score is 45.7, high compared with the total Value average of 39.6.

Quality Score50.6coverage 91.2%: missing Epoch General ECI, CursorBench 3.2 and DeepSWE v1.1
Value Score45.7
Reliability45.1
Cache Discount99%

Model specification

Reasoning
reasoning
Input modalities
none
Output modalities
none
Context window
not available
Weights
open weights

Token prices USD per 1M tokens

Input
$0.435
Output
$0.870
Cached Input
$0.0036
Cache write
not available
Cost per task
$0.040

Capability

Intelligence
43.1
Coding
58.7
Agentic
34.4
Omniscience
-9.7
Correct
41.8%
Blended price
$0.177

Answer outcomes

Attempt rate: 93.4%

Correct 41.8%Incorrect 51.5%Abstained 6.6%

Artificial Analysis benchmarks

GDPval-AA v2
40%
τ³-Banking
24%
Terminal-Bench v2.1
65%
SciCode
46%
Humanity’s Last Exam
34%
GPQA Diamond
91%
CritPt
10%
AA-Omniscience
45%
AA-LCR
65%

Similar models