Navigasi
AI BENCHY
Advertise here

AI BENCHY Compare

Model yang Dibandingkan

Ringkasan

Perbandingan benchmark DeepSeek V4 Pro vs DeepSeek V4 Flash vs DeepSeek V3.2 vs Kimi K2.6DeepSeek V4 Flash unggul pada Skor dengan 8.3. DeepSeek V4 Flash unggul pada Keandalan dengan 10.0. DeepSeek V4 Flash memiliki Total Biaya terendah di $0.029. DeepSeek V4 Flash paling cepat di 45.85s.

Model yang direkomendasikan: DeepSeek V4 Flash - It has the best score here (8.3), while costing about 11.8x less than model lain dalam perbandingan ini.

Benchmark dihasilkan dari suite pengujian AI BENCHY pada: 2026-06-12

Metrik DeepSeek V4 Pro DeepSeek V4 Pro high Rilis: 2026-04-24 DeepSeek V4 Flash DeepSeek V4 Flash high Rilis: 2026-04-24 DeepSeek V3.2 DeepSeek V3.2 medium Rilis: 2025-12-01 Kimi K2.6 Kimi K2.6 medium Rilis: 2026-04-20 Tersedia gratis
Skor 6.3 8.3 7.5 7.8
Peringkat #90 #26 #50 #38
Keandalan 9.0 10.0 10.0 10.0
Konsistensi 7.6 8.5 7.6 8.6
Tes benar
Tingkat lulus per percobaan 52.4% 74.6% 65.1% 65.1%
Tes tidak stabil 6 4 6 3
Total Run 63 63 63 63
Biaya per hasil 2.869 0.299 0.436 8.358
Total Biaya $0.079 $0.029 $0.044 $0.889
Harga input $0.435 / 1M $0.098 / 1M $0.229 / 1M $0.680 / 1M
Harga output $0.870 / 1M $0.196 / 1M $0.344 / 1M $3.410 / 1M
Total token input 32,240 39,745 38,333 29,450
Token output 12,250 10,310 7,186 102,923
Token penalaran 72,257 123,501 99,081 254,094
Waktu respons (rata-rata) 65.21s 45.85s 68.71s 71.67s
Waktu respons (maks) 358.35s 218.13s 376.10s 406.78s
Waktu respons (total) 1304.19s 962.79s 1442.81s 1433.36s

Generation showcase

Hamster playing table tennis

Prompt: Create a detailed SVG illustration of a hamster playing table tennis.

#90 DeepSeek V4 Pro

high
Cost
$0.023
Time
257.6s
Tokens
14,870 tok

#26 DeepSeek V4 Flash

high
Cost
$0.003
Time
93.1s
Tokens
7,926 tok

#50 DeepSeek V3.2

medium
Cost
$0.001
Time
53.6s
Tokens
1,932 tok

#38 MoonshotAI: Kimi K2.6

medium
Cost
$0.013
Time
103.4s
Tokens
3,620 tok

Model teratas berdasarkan skor

Skor vs Total Biaya

Waktu respons (rata-rata)

Skor vs Waktu respons (rata-rata)

Total token output

Skor vs Total token output

Rincian Kategori

Trik anti-AI Skor Konsistensi Tingkat lulus per percobaan Tes tidak stabil Tes benar Waktu respons (rata-rata) Token input Token output Token penalaran
DeepSeek V4 Pro 6.4 7.9 58.3% 1 16.53s 448 71 3,617
DeepSeek V4 Flash 8.3 10.0 75.0% 0 28.51s 540 140 7,770
DeepSeek V3.2 8.2 7.9 83.3% 1 24.23s 448 3,247 6,953
Kimi K2.6 7.0 8.0 66.7% 1 11.59s 618 7,115 8,934
Pemrograman Skor Konsistensi Tingkat lulus per percobaan Tes tidak stabil Tes benar Waktu respons (rata-rata) Token input Token output Token penalaran
DeepSeek V4 Pro 3.3 6.4 11.1% 1 118.23s 1,966 111 20,940
DeepSeek V4 Flash 7.8 10.0 66.7% 0 50.60s 7,279 395 34,862
DeepSeek V3.2 6.0 7.2 55.6% 1 248.68s 5,717 649 52,014
Kimi K2.6 5.7 8.6 33.3% 0 214.42s 2,925 9,970 77,189
Gabungan Skor Konsistensi Tingkat lulus per percobaan Tes tidak stabil Tes benar Waktu respons (rata-rata) Token input Token output Token penalaran
DeepSeek V4 Pro 10.0 10.0 100.0% 0 65.02s 14,016 465 5,914
DeepSeek V4 Flash 10.0 10.0 100.0% 0 76.57s 14,016 465 7,347
DeepSeek V3.2 10.0 10.0 100.0% 0 93.11s 14,283 571 6,296
Kimi K2.6 10.0 10.0 100.0% 0 40.96s 11,271 711 13,876
Parsing dan ekstraksi data Skor Konsistensi Tingkat lulus per percobaan Tes tidak stabil Tes benar Waktu respons (rata-rata) Token input Token output Token penalaran
DeepSeek V4 Pro 7.3 5.9 83.3% 1 23.62s 5,633 229 1,710
DeepSeek V4 Flash 10.0 10.0 100.0% 0 28.03s 7,290 201 1,179
DeepSeek V3.2 10.0 10.0 100.0% 0 36.09s 7,388 207 7,693
Kimi K2.6 10.0 10.0 100.0% 0 20.38s 7,014 316 11,305
Spesifik domain Skor Konsistensi Tingkat lulus per percobaan Tes tidak stabil Tes benar Waktu respons (rata-rata) Token input Token output Token penalaran
DeepSeek V4 Pro 2.9 7.2 11.1% 1 205.66s 430 10,529 28,089
DeepSeek V4 Flash 4.1 4.4 44.5% 2 100.31s 666 27 59,249
DeepSeek V3.2 2.9 4.4 22.2% 2 24.27s 472 21 6,838
Kimi K2.6 5.3 7.2 44.4% 1 202.38s 326 47,035 98,262
Kecerdasan umum Skor Konsistensi Tingkat lulus per percobaan Tes tidak stabil Tes benar Waktu respons (rata-rata) Token input Token output Token penalaran
DeepSeek V4 Pro 6.1 3.1 66.7% 1 25.09s 314 76 1,152
DeepSeek V4 Flash 6.1 3.1 66.7% 1 25.15s 471 79 632
DeepSeek V3.2 3.4 2.5 33.3% 1 58.29s 314 49 2,189
Kimi K2.6 10.0 10.0 100.0% 0 17.83s 477 3,981 4,472
Kepatuhan instruksi Skor Konsistensi Tingkat lulus per percobaan Tes tidak stabil Tes benar Waktu respons (rata-rata) Token input Token output Token penalaran
DeepSeek V4 Pro 10.0 10.0 100.0% 0 41.16s 627 205 2,416
DeepSeek V4 Flash 10.0 10.0 100.0% 0 15.36s 627 63 1,622
DeepSeek V3.2 10.0 10.0 100.0% 0 35.78s 627 1,397 2,845
Kimi K2.6 10.0 10.0 100.0% 0 12.53s 669 3,977 5,269
Pemecahan teka-teki Skor Konsistensi Tingkat lulus per percobaan Tes tidak stabil Tes benar Waktu respons (rata-rata) Token input Token output Token penalaran
DeepSeek V4 Pro 5.9 7.2 55.6% 1 34.84s 544 139 4,019
DeepSeek V4 Flash 8.2 7.2 88.9% 1 26.11s 594 196 1,767
DeepSeek V3.2 7.0 7.2 55.6% 1 37.69s 594 518 6,375
Kimi K2.6 6.0 7.4 55.6% 1 25.06s 651 13,860 17,599
Pemanggilan alat Skor Konsistensi Tingkat lulus per percobaan Tes tidak stabil Tes benar Waktu respons (rata-rata) Token input Token output Token penalaran
DeepSeek V4 Pro 10.0 10.0 100.0% 0 21.33s 8,079 372 593
DeepSeek V4 Flash 10.0 10.0 100.0% 0 74.73s 8,079 228 542
DeepSeek V3.2 10.0 10.0 100.0% 0 34.81s 8,307 507 859
Kimi K2.6 10.0 10.0 100.0% 0 8.92s 5,286 248 1,011
Pengetahuan umum Skor Konsistensi Tingkat lulus per percobaan Tes tidak stabil Tes benar Waktu respons (rata-rata) Token input Token output Token penalaran
DeepSeek V4 Pro 3.0 10.0 0.0% 0 39.14s 183 53 3,807
DeepSeek V4 Flash 3.0 10.0 0.0% 0 54.46s 183 8,516 8,531
DeepSeek V3.2 3.0 10.0 0.0% 0 83.99s 183 20 7,019
Kimi K2.6 3.0 10.0 0.0% 0 130.27s 213 15,710 16,177

Perbandingan Cepat

Ganti Pasangan Perbandingan

DeepSeek V4 ProhighvsGPT-5.5noneDeepSeek V4 ProhighvsQwen3.5-35B-A3BmediumDeepSeek V4 FlashhighvsGemini 2.5 FlashmediumDeepSeek V4 ProhighvsGemma 4 31BmediumTersedia gratisDeepSeek V3.2mediumvsGPT-5.3 ChatnoneDeepSeek V4 ProhighvsNemotron 3 SupermediumTersedia gratisKimi K2.6mediumTersedia gratisvsStep 3.7 FlashlowSeed-2.0-LitenonevsDeepSeek V4 ProhighDeepSeek V3.2mediumvsGemini 3 Flash PreviewlowDeepSeek V4 FlashhighvsQwen3.7 PlusmediumDeepSeek V4 ProhighvsGemini 2.5 FlashnoneDeepSeek V4 FlashhighvsGPT-5.2medium