GaMS3-12B-Instruct
CJVTodprte uteži
- Licenca
- Gemma Terms of Use
- ID
cjvt/GaMS3-12B-Instruct
GaMS3-12B-Instruct ima v Slovenski pogovorni areni oceno 1.060 (±55), ki jo je določilo 392 glasov v slepih primerjavah v slovenščini. 95-odstotni interval ocene je 1.005–1.115: pove, kako natančno so jo glasovi doslej določili. Med odločenimi primerjavami zmaga v 61,2 % (172 zmag, 109 porazov, 111 izenačenj). Mediana dolžine njegovih odgovorov je 201 beseda. Razvijalec modela je CJVT. Model ima odprte uteži (licenca: Gemma Terms of Use).
Na kratko
- Ocena
- 1.060±55
- 95-odstotni interval 1.005–1.115
- Dolžina odgovora
- 201beseda
- mediana, običajno 33–525
- Glasovi
- 392
- 61,2 % zmag med odločenimi
Informacije o modelu
GaMS (Generativni model za slovenščino) Centra za jezikovne vire in tehnologije Univerze v Ljubljani (CJVT). Model za sledenje navodilom, zgrajen na osnovi modela Google Gemma 3 12B.
Statistika
Ocena s 95-odstotnim intervalom zaupanja
- GaMS2-27B-Instruct
- GaMS-27B-Instruct-Nemotronni aktiven
- gpt-5
- GaMS3-12B-Instruct
- gemini-2.0-flashni aktiven
- Mistral-Large-3-675B-Instruct-2512ni aktiven
- gemma-3-27b-itni aktiven
Modeli tik nad njim in pod njim na lestvici, na isti osi. Navpična oznaka je ocena, črta 95-odstotni interval.
Izidi glasovanj
- Zmage172
- Izenačeno111
- Porazi109
- Število poizvedb
- 601
Dolžina odgovora med modeli
Mediana besed na odgovor za vsak model na lestvici, na isti osi.
Daljši odgovori kot pri 24 od 39 drugih modelov.
Vsi modeli v številkah
| Model | Mediana besed |
|---|---|
| gpt-oss-120b | 463 |
| Ministral-3-14B-Instruct-2512 | 348 |
| deepseek-r1 | 336 |
| gemini-2.5-pro | 330 |
| Qwen3-Next-80B-A3B-Instruct | 312 |
| gemma-4-26b-a4b-it | 306 |
| Qwen3.6-27B | 284 |
| gpt-oss-20b | 284 |
| gemini-3-pro-preview | 260 |
| gemini-2.5-flash | 242 |
| GaMS-9B-Instruct-Nemotron | 229 |
| GaMS-27B-Instruct-Nemotron | 226 |
| gemini-3.1-pro-preview | 224 |
| Mistral-Large-3-675B-Instruct-2512 | 224 |
| deepseek-r1-0528 | 216 |
| GaMS3-12B-Instruct | 201 |
| nemotron-3-ultra-550b-a55b | 199 |
| GaMS2-27B-Instruct | 196 |
| gemma-4-31b-it | 179 |
| gemini-3.8-flash | 178 |
| llama3-70b-instruct | 168 |
| gpt-5-mini | 152 |
| kimi-k2-instruct | 142 |
| gemini-3.5-flash | 141 |
| Mistral-Small-24B-Instruct-2501 | 110 |
| gemma-3-12b-it | 107 |
| bielik-11b-v2.3-instruct | 104 |
| gpt-5 | 101 |
| gemma-3-27b-it | 101 |
| gemini-1.5-pro | 100 |
| llama-4-scout-17b-16e-instruct | 91 |
| llama-3.3-70b-instruct | 90 |
| EuroLLM-9B-Instruct | 83 |
| gemini-2.0-flash | 78 |
| llama-4-maverick-17b-128e-instruct | 78 |
| gpt-5.6-luna | 77 |
| gpt-4o | 68 |
| Zlatorog-12B-Instruct-Beta | 47 |
| GaMS-9B-Instruct | 13 |
| GaMS-27B-Instruct | 10 |
Dolžina odgovora v žetonih
Delež odgovorov po številu žetonov, iz 56 odgovorov.
Žeton je delček besede, kot ga šteje model. Razmišljanje modela ni šteto.
Podatki po razredih
| Žetoni | Odgovori | Delež |
|---|---|---|
| 0–128 | 13 | 23 % |
| 128–256 | 9 | 16 % |
| 256–384 | 3 | 5 % |
| 384–512 | 3 | 5 % |
| 512–640 | 3 | 5 % |
| 640–768 | 3 | 5 % |
| 768–896 | 2 | 4 % |
| 896–1.024 | 5 | 9 % |
| 1.024–1.152 | 3 | 5 % |
| 1.152–1.280 | 6 | 11 % |
| 1.280–1.408 | 0 | 0 % |
| 1.408–1.536 | 3 | 5 % |
| 1.536–1.664 | 0 | 0 % |
| 1.664–1.792 | 0 | 0 % |
| 1.792–1.920 | 2 | 4 % |
| 1.920–2.048 | 0 | 0 % |
| 2.048–2.176 | 1 | 2 % |
| 2.176–2.304 | 0 | 0 % |
| 2.304–2.432 | 0 | 0 % |
| 2.432–2.560 | 0 | 0 % |
| 2.560–2.688 | 0 | 0 % |
| 2.688–2.816 | 0 | 0 % |
| 2.816–2.944 | 0 | 0 % |
| 2.944–3.072 | 0 | 0 % |
| 3.072–3.200 | 0 | 0 % |
| 3.200–3.328 | 0 | 0 % |
| 3.328–3.456 | 0 | 0 % |
| 3.456–3.584 | 0 | 0 % |
| 3.584–3.712 | 0 | 0 % |
| 3.712–3.840 | 0 | 0 % |
| 3.840–3.968 | 0 | 0 % |
| 3.968–4.096 | 0 | 0 % |
| 4.096–4.224 | 0 | 0 % |
| 4.224–4.352 | 0 | 0 % |
| 4.352–4.480 | 0 | 0 % |
| 4.480–4.608 | 0 | 0 % |
| 4.608–4.736 | 0 | 0 % |
| 4.736–4.864 | 0 | 0 % |
| 4.864–4.992 | 0 | 0 % |
| 4.992–5.120 | 0 | 0 % |
| 5.120–5.248 | 0 | 0 % |
| 5.248–5.376 | 0 | 0 % |
| 5.376–5.504 | 0 | 0 % |
| 5.504–5.632 | 0 | 0 % |
| 5.632–5.760 | 0 | 0 % |
| 5.760–5.888 | 0 | 0 % |
| 5.888–6.016 | 0 | 0 % |
| 6.016–6.144 | 0 | 0 % |
| 6.144–6.272 | 0 | 0 % |
| 6.272–6.400 | 0 | 0 % |
| 6.400–6.528 | 0 | 0 % |
| 6.528–6.656 | 0 | 0 % |
| 6.656–6.784 | 0 | 0 % |
| 6.784–6.912 | 0 | 0 % |
| 6.912–7.040 | 0 | 0 % |
| 7.040–7.168 | 0 | 0 % |
| 7.168–7.296 | 0 | 0 % |
| 7.296–7.424 | 0 | 0 % |
| 7.424–7.552 | 0 | 0 % |
| 7.552–7.680 | 0 | 0 % |
| 7.680–7.808 | 0 | 0 % |
| 7.808–7.936 | 0 | 0 % |
| 7.936–8.064 | 0 | 0 % |
| 8.064–8.192 | 0 | 0 % |