readable AI benchmarks (simplified)

readable AI benchmarks

StepFun

Step 3.5 Flash 2603

reasoning · closed weights · Apr 2, 2026

Step 3.5 Flash 2603 has a Quality Score of 27.0, ranking 284th among 551 scored models. Per 1M tokens, pricing is $0.100 input, $0.300 output, and $0.020 cached input. It accepts text input, outputs text, and has a 256k token context window. Its Reliability Score is 27.9, ranking 335th among 553 scored models, below average (total average: 35.6). Its Value Score is 25.0, low compared with the total Value average of 30.7.

Quality Score27.0coverage 100.0%: AA Intelligence and AA-Omniscience outcomes available
Value Score25.0
Factual reliability27.9
Cache Discount80%

Model specification

Reasoning
reasoning
Input modalities
text
Output modalities
text
Context window
256k tokens
Weights
closed weights

Token prices USD per 1M tokens

Input
$0.100
Output
$0.300
Cached Input
$0.020
Cache write
not available
Cost per task
not available

Capability

Intelligence
17.0
Coding
not available
Agentic
not available
Omniscience
-44.2
Correct
24.6%
Blended price
$0.064

Answer outcomes

Fully graded outcomes (Correct + Incorrect): 93.5%

Correct 24.6%Incorrect 68.8%Partial / not attempted 6.6%

Artificial Analysis benchmarks

Humanity’s Last Exam
24%
GPQA Diamond
83%
CritPt
2%
AA-Omniscience
28%
AA-LCR
63%

Similar models

Support me! Patreon