readable AI benchmarks (simplified)

readable AI benchmarks

Sapiens AI

Agnes 2.5 Pro Beta

reasoning · closed weights · Aug 26, 2026

Agnes 2.5 Pro Beta has a Quality Score of 44.8, ranking 60th among 263 scored models. Per 1M tokens, pricing is $0.100 input, $0.300 output, and $0.010 cached input. It accepts text and image input, outputs text, and has a 1M token context window. Its Reliability Score is 44.7, ranking 71st among 263 scored models, above average (total average: 33.3). Its Value Score is 47.9, high compared with the total Value average of 27.5.

Quality Score44.8coverage 91.2%: missing Epoch General ECI, CursorBench 3.2 and DeepSWE v1.1
Value Score47.9
Reliability44.7
Cache Discount90%

Model specification

Reasoning
reasoning
Input modalities
text, image
Output modalities
text
Context window
1M tokens
Weights
closed weights

Token prices USD per 1M tokens

Input
$0.100
Output
$0.300
Cached Input
$0.010
Cached Output
not available
Cost per task
not available

Capability

Intelligence
49.1
Coding
62.3
Agentic
43.8
Omniscience
-10.5
Correct
16.8%
Blended price
$0.057

Answer outcomes

Attempt rate: 44.1%

Correct 16.8%Incorrect 27.3%Abstained 55.9%

Artificial Analysis benchmarks

GDPval-AA v2
48%
τ³-Banking
36%
Terminal-Bench v2.1
70%
SciCode
48%
Humanity’s Last Exam
38%
GPQA Diamond
91%
CritPt
16%
AA-Omniscience
45%
AA-LCR
78%

Similar models