Models
Parameter counts come from the model card linked on each page. Memory on those pages is calculated, not copied from a benchmark.
Llama
gpt-oss
- gpt-oss 20B20.9B · 3.6B active
- gpt-oss 120B117B · 5.1B active
Qwen
- Qwen3 8B8B
- Qwen3 14B14B
- Qwen3 30B-A3B30B · 3B active
- Qwen3 235B-A22B235B · 22B active
- Qwen3.8 27B27B
- Qwen3.6 35B-A3B35B · 3B active
Gemma
- Gemma 3 12B12B
- Gemma 4 E4B8B
- Gemma 4 12B12B
- Gemma 4 26B-A4B25.2B · 3.8B active
- Gemma 4 31B30.7B
DeepSeek
- DeepSeek R1 Distill 8B8B
- DeepSeek R1 Distill 14B14B
- DeepSeek R1 Distill 70B70B
- DeepSeek R1 671B671B · 37B active