#30
DeepSeek V3.2
Deepseek ยท Toleo: 2025-12-01 ยท deepseek/deepseek-v3.2::none
Wastani wa alama
4.70
Gharama kwa matokeo
0.132
Uthabiti
8.20
Jumla ya gharama
$0.007
Majaribio sahihi
5
Jaribio huhesabiwa kuwa limepita kikamilifu tu ikiwa run zake zote zimepita.
Majaribio yenye makosa
9
Kiwango cha kupita kwa kila jaribio: 47.6%
Majaribio yasiyo thabiti
3
Muda wa majibu: wastani 6.22s ยท jumla 87.03s ยท upeo 16.20s
Jibu lisilo sahihi: 6
Muundo wa ziada: 2
Hakufuata maelekezo: 1
Modeli bora kwa alama
Chagua modeli ya kwanza, kisha bofya modeli ya pili kufungua ukurasa wa kulinganisha bega kwa bega.
Ulinganisho wa haraka
DeepSeek V3.2nonevsGPT-4o-mininoneDeepSeek V3.2nonevsQwen3.5-27BnoneDeepSeek V3.2nonevsQwen3.5-122B-A10BnoneDeepSeek V3.2nonevsQwen3.5-35B-A3BnoneDeepSeek V3.2nonevsMiniMax M2.5mediumDeepSeek V3.2nonevsTrinity Large Preview (free)noneInapatikana bureDeepSeek V3.2nonevsGemini 3 Flash PreviewmediumDeepSeek V3.2nonevsGemini 3.1 Pro PreviewmediumDeepSeek V3.2nonevsStep 3.5 FlashmediumInapatikana bure
Mgawanyo wa kategoria
| Kategoria | Wastani wa alama | Uthabiti | Majaribio sahihi |
|---|---|---|---|
| Anti-AI Tricks | 1.00 | 9.68 | 0/3 |
| Data parsing and extraction | 5.38 | 5.81 | 1/2 |
| Domain specific | 1.00 | 7.21 | 0/3 |
| Instructions following | 8.00 | 9.99 | 1/2 |
| Puzzle Solving | 7.67 | 7.50 | 2/3 |
| Tool Calling | 10.00 | 10.00 | 1/1 |