- Intelligence
- 3.8
- Rank 778 of 782 measured models
- Coding
- 5.8
- Rank 286 of 301 measured models
- Output speed
- 52.6 t/s
- Rank 364 of 449 measured models
- Context window
- 128K
- Rank 845 of 1325 measured models
- Cost per task
- $0
- Rank 1 of 229 measured models
Gemma 3 12B Instruct scores 3.8 on the Artificial Analysis Intelligence Index, ranking 778th of 782 measured models. The closest models on this index are Mistral Small 4 (4.3), gpt-oss-120b (3.8), and North Mini Code (3.4). With an output speed of 52.6 t/s, it is slower than Mistral Small 4 (143 t/s), gpt-oss-120b (163 t/s), and North Mini Code (190 t/s).
Models at a similar level
The closest models on the Intelligence Index, one configuration per family.
- Gemma 4 31B
Google · — · 23.6 t/s
4.4+0.6—23.6 t/sCompare - Mistral Small 4
Mistral · $0.262 · 143 t/s
4.3+0.5$0.262143 t/sCompare - gpt-oss-120b
OpenAI · $0.262 · 163 t/s
3.8$0.262163 t/sCompare - North Mini Code
Cohere · — · 190 t/s
3.4−0.4—190 t/sCompare
