Model benchmark rankings

Compare capability scores, evaluation time, and reference cost across models and reasoning efforts.

Evaluated on private datasetsWhy we test this way

Data snapshot · 10/27/2026

Scoring dimension
About the scores

Service logic, error handling, and test verification. By default, each model shows its highest-scoring tested effort for the selected dimension.

Evaluation time and reference cost correspond to the selected dimension. They reflect benchmark runs, not everyday single-request latency or billing.

Explore evaluation tasksEvaluation method
Filters
Result scope
44 models · 105 configsBest effort per model

Scroll horizontally to view all scores and costs →

RankModelEval. time
Speed tok/sActions
01
gpt-6-astra
Backend engineering97Frontend interaction94Knowledge & Reasoning95Visual aesthetics73Maintainability61
88.50979495736145m 41s$4.80…
02
gpt-6.1-sol
Backend engineering95Frontend interaction96Knowledge & Reasoning90Visual aesthetics66Maintainability70
86.60959690667032m 36s$1.06…
03
gpt-6.1-terraNEW
Backend engineering89Frontend interaction84Knowledge & Reasoning85Visual aesthetics64Maintainability61
80.0589848564617m 56s$0.38…
04
grok-4.7
Backend engineering87Frontend interaction80Knowledge & Reasoning85Visual aesthetics66Maintainability55
78.55878085665577m 29s$2.00…
05
claude-sonnet-5-5
Backend engineering86Frontend interaction89Knowledge & Reasoning85Visual aesthetics73Maintainability66
82.30868985736615m 3s$1.50…
06
claude-fable-5-1
Backend engineering85Frontend interaction91Knowledge & Reasoning95Visual aesthetics80Maintainability68
87.25859195806843m 45s$12.77…
06
gpt-5.6-sol
Backend engineering85Frontend interaction73Knowledge & Reasoning75Visual aesthetics66Maintainability64
74.15857375666429m 19s$1.71…
08
claude-opus-5-5
Backend engineering84Frontend interaction97Knowledge & Reasoning100Visual aesthetics73Maintainability70
89.15849710073709m 29s$1.05…
08
k3.1-proNEW
Backend engineering84Frontend interaction91Knowledge & Reasoning80Visual aesthetics68Maintainability64
79.60849180686442m 34s$1.8784
10
Gemini 4 Argon
Backend engineering82Frontend interaction84Knowledge & Reasoning85Visual aesthetics70Maintainability59
79.35828485705914m 48sN/A…
10
muse-wonder
Backend engineering82Frontend interaction84Knowledge & Reasoning90Visual aesthetics73Maintainability59
81.55828490735914m 35s$2.53…
12
gpt-5.6-terra
Backend engineering81Frontend interaction76Knowledge & Reasoning55Visual aesthetics61Maintainability61
65.90817655616175m 45s$0.70…
13
gpt-6-sol
Backend engineering80Frontend interaction94Knowledge & Reasoning75Visual aesthetics75Maintainability66
78.90809475756615m 13s$0.46…
14
grok-4.6
Backend engineering78Frontend interaction80Knowledge & Reasoning85Visual aesthetics64Maintainability59
76.85788085645943m 8s$0.71…
15
MiniMax-M3.1-Flash-Preview
Backend engineering75Frontend interaction57Knowledge & Reasoning75Visual aesthetics70Maintainability53
68.45755775705318m 35sN/A…
16
claude-opus-5
Backend engineering74Frontend interaction87Knowledge & Reasoning85Visual aesthetics66Maintainability59
77.75748785665918m 42s$2.01…
16
hy4-preview
Backend engineering74Frontend interaction73Knowledge & Reasoning65Visual aesthetics59Maintainability57
66.70747365595767m 39s$0.40…
18
MiniMax-M3.1-FlashNEW
Backend engineering73Frontend interaction75Knowledge & Reasoning80Visual aesthetics66Maintainability61
73.60737580666130m 52s$0.11…
18
muse-spark-1.3
Backend engineering73Frontend interaction91Knowledge & Reasoning75Visual aesthetics59Maintainability47
72.6073917559476m 23s$0.024…
20
deepseek-v4.1-flash-1017NEW
Backend engineering72Frontend interaction96Knowledge & Reasoning75Visual aesthetics61Maintainability53
74.30729675615315m 16s$0.12…
21
glm-5.3
Backend engineering70Frontend interaction78Knowledge & Reasoning70Visual aesthetics66Maintainability64
70.40707870666417m 14s$0.44…
21
k3
Backend engineering70Frontend interaction85Knowledge & Reasoning70Visual aesthetics68Maintainability59
71.60708570685937m 56s$0.89…
21
mimo-v2.6-pro
Backend engineering70Frontend interaction85Knowledge & Reasoning65Visual aesthetics64Maintainability59
69.25708565645980m 5s$0.18…
24
claude-haiku-5-5
Backend engineering69Frontend interaction78Knowledge & Reasoning60Visual aesthetics59Maintainability64
65.6569786059648m 11s$0.077…
25
doubao-seed-2-1-pro
Backend engineering68Frontend interaction68Knowledge & Reasoning55Visual aesthetics68Maintainability49
61.55686855684919m 43s$0.45…
25
Qwen 3.8 Flash
Backend engineering68Frontend interaction87Knowledge & Reasoning50Visual aesthetics73Maintainability57
65.15688750735755m 14s$0.12…
27
deepseek-v4-flash-vision-exp
Backend engineering67Frontend interaction15Knowledge & Reasoning65Visual aesthetics—Maintainability—
—671565——14m 12s$0.11…
27
deepseek-v4-pro
Backend engineering67Frontend interaction63Knowledge & Reasoning65Visual aesthetics70Maintainability59
65.15676365705926m 53s$0.44…
29
gpt-5.6-luna
Backend engineering65Frontend interaction78Knowledge & Reasoning25Visual aesthetics59Maintainability61
52.30657825596148m 26s$0.17…
30
claude-opus-4-8
Backend engineering64Frontend interaction88Knowledge & Reasoning85Visual aesthetics68Maintainability55
75.85648885685563m 16s$8.37…
30
Gemini 3.8 Flash High
Backend engineering64Frontend interaction86Knowledge & Reasoning40Visual aesthetics66Maintainability59
59.80648640665913m 35s$0.061…
30
Qwen 3.8 27b
Backend engineering64Frontend interaction87Knowledge & Reasoning20Visual aesthetics66Maintainability59
53.00648720665922m 59s$0.43…
30
Qwen 3.8 Max
Backend engineering64Frontend interaction69Knowledge & Reasoning40Visual aesthetics64Maintainability53
55.50646940645347m 48s$1.07…
34
Gemini 3.7 Flash High
Backend engineering63Frontend interaction71Knowledge & Reasoning25Visual aesthetics70Maintainability59
51.9563712570595m 56s$0.069…
35
deepseek-v4.1-flash
Backend engineering62Frontend interaction88Knowledge & Reasoning65Visual aesthetics64Maintainability55
67.85628865645512m 33s$0.10…
35
gpt-6-luna
Backend engineering62Frontend interaction80Knowledge & Reasoning50Visual aesthetics68Maintainability59
62.00628050685917m 54s$0.023…
37
space-bunny
Backend engineering61Frontend interaction80Knowledge & Reasoning75Visual aesthetics59Maintainability55
68.80618075595519m 33s$0.000…
37
step-5
Backend engineering61Frontend interaction71Knowledge & Reasoning75Visual aesthetics47Maintainability55
65.2061717547557m 12s$0.34…
39
deepseek-v4-flash
Backend engineering58Frontend interaction76Knowledge & Reasoning55Visual aesthetics61Maintainability53
60.50587655615311m 36s$0.090…
39
mimo-v2.6-flash
Backend engineering58Frontend interaction71Knowledge & Reasoning45Visual aesthetics59Maintainability57
56.10587145595735m 36s$0.043…
41
glm-5.3-flash
Backend engineering55Frontend interaction76Knowledge & Reasoning20Visual aesthetics70Maintainability59
49.60557620705921m 43s$0.018…
42
k2.8
Backend engineering54Frontend interaction72Knowledge & Reasoning70Visual aesthetics57Maintainability53
63.55547270575322m 14s$0.21…
43
MiniMax-M3
Backend engineering46Frontend interaction65Knowledge & Reasoning50Visual aesthetics59Maintainability44
52.95466550594422m 6s$0.31…
43
Nemotron-3-Ultra
Backend engineering46Frontend interaction40Knowledge & Reasoning40Visual aesthetics61Maintainability37
44.05464040613717m 3s$0.000…
Each model uses its highest-scoring effort here. Tied scores share a rank.
Service logic, error handling, and test verification. Evaluation scope and sources