#319 Step 3.5 Flash
medium- Gharama
- $0.008
- Muda
- 277.1s
- Tokeni
- 23,695 tok
Stepfun
Toleo: 2026-02-01
Imepimwa tarehe: 2026-10-01 02:07
stepfun/step-3.5-flash::medium
196B jumla (11B vinavyotumika)MoEChanzo huria
(medium)
(none)
Muhtasari
Step 3.5 Flash hupata alama 5.4 kwenye AI BENCHY na iko nafasi ya #319. Ina reliability 10.0, pass rate 49.3%, gharama jumla $0.105, na wastani wa response time 190.48s.
Kinachofanya Step 3.5 Flash iwe ya kipekee: Inaonekana zaidi kwenye Mbinu za kupinga AI, ambako iko #1; huku Kazi za mawakala ikiwa eneo lake dhaifu zaidi kwenye #13.
Taarifa za modeli
Ilifanyiwa utafiti tarehe 2026-08-12
- Vigezo
- 196B jumla (11B vinavyotumika)
- Muundo
- MoE
- Upatikanaji
- Chanzo huria
- Leseni
- Apache-2.0
5.4
Uthabiti
8.3
10.0
Jumla ya tokeni za matokeo
454,482
Jumla ya tokeni za ingizo
139,086
Bei ya ingizo
$0.100 / 1M
Bei ya toleo
$0.300 / 1M
Bei ya kusoma akiba
Haipo
Bei ya kuandika akiba
Haipo
Bei za akiba zinahusu tokeni za ingizo. Kusoma hutumia tena maelekezo yaliyohifadhiwa; kuandika huyahifadhi na kunaweza kuongeza gharama. Tokeni za matokeo hutumia bei ya matokeo.
Majaribio yenye makosa: 12
Kiwango cha kupita kwa kila jaribio: 49.3%
Ufunikaji wa benchmark: marudio 66/69 (Lengo: marudio 3 kwa kila jaribio)
Majaribio yasiyo thabiti
3
Majaribio yasiyo thabiti yalikuwa na matokeo mchanganyiko kati ya run (angalau kupita moja na kufeli moja).
Hamster playing table tennis
Prompt: Create a detailed SVG illustration of a hamster playing table tennis.
Historia ya uendeshaji
| Imepimwa tarehe | Alama | Uaminifu | Majaribio sahihi | Jumla ya gharama | Linganisha |
|---|---|---|---|---|---|
| 2026-10-01 02:07 Jaribio tena | 5.4 | 10.0 | $0.105 | Uendeshaji wa sasa | |
| 2026-07-16 23:25 Jaribio jipya limeongezwa | 6.0 | 9.2 | $0.108 ↑ | Linganisha | |
| 2026-06-04 14:12 Jaribio jipya limeongezwa | 7.2 | 8.9 | $0.070 ↑ | Linganisha | |
| 2026-05-22 00:30 Suite imebadilika | 7.4 | 9.3 | $0.015 | Linganisha | |
| 2026-05-08 15:30 Suite imebadilika | 7.6 | 10.0 | $0.011 | Linganisha | |
| 2026-04-11 00:35 Jaribio la kwanza lililorekodiwa | 7.9 | Haipo | $0.000 | Linganisha |
Uendeshaji huu ulitumia suite tofauti ya benchmark. Zingatia mabadiliko ya suite unapochambua mwenendo wa kihistoria.
Chati
Chagua modeli ya kwanza, kisha bofya modeli ya pili kufungua ukurasa wa kulinganisha bega kwa bega.
Modeli bora kwa alama
Alama vs Jumla ya gharama
Muda wa majibu (wastani)
Alama vs Muda wa majibu (wastani)
Jumla ya tokeni za matokeo
Alama vs Jumla ya tokeni za matokeo
Ulinganisho wa haraka
Step 3.5 FlashmediumvsMiMo-V2.6-FlashnoneStep 3.5 FlashmediumvsEmber-1noneStep 3.5 FlashmediumvsHy4 previewlowStep 3.5 FlashmediumvsGPT-5.6 LunanoneStep 3.5 FlashmediumvsNorth Mini CodemediumInapatikana bureStep 3.5 FlashmediumvsGranite 4.2 8BlowStep 3.5 FlashmediumvsQwen3.8 27BnoneInapatikana bureStep 3.5 FlashmediumvsKimi K2.5noneStep 3.5 FlashmediumvsSeed-2.0-CodenoneStep 3.5 FlashmediumvsLaguna XS 2.1noneInapatikana bure
Mgawanyo wa kategoria
| Kategoria | Alama | Uthabiti | Majaribio sahihi |
|---|---|---|---|
| Kazi za mawakala | 2.8 | 1.6 | |
| Mbinu za kupinga AI | 10.0 | 10.0 | |
| Uandishi wa msimbo | 2.4 | 5.2 | |
| Mchanganyiko | 6.5 | 10.0 | |
| Uchanganuzi na uchimbaji wa data | 10.0 | 10.0 | |
| Mahususi kwa domeni | 3.5 | 4.4 | |
| Akili ya jumla | 5.5 | 10.0 | |
| Ufuataji wa maagizo | 8.3 | 10.0 | |
| Utatuzi wa mafumbo | 5.3 | 10.0 | |
| Mwito wa zana | 10.0 | 10.0 | |
| Maarifa ya jumla | 3.0 | 10.0 |