STSentence Transformersall-minilmLightweight embedding modelMinimum VRAM0.3 GB0.3-0.5 GB comfort rangeModel scale0.07B256 contextBest forEmbeddingsCommunity fitCustomQ4_K_MOpen →
COCohereAya 23 35BLarger multilingual model from CohereMinimum VRAM17 GB17-26 GB comfort rangeModel scale35B8,192 contextBest forGeneral ChatCommunity fitCustomQ4_K_MOpen →
COCohereAya 23 8BCohere's multilingual model supporting 23 languagesMinimum VRAM5.5 GB5.5-10 GB comfort rangeModel scale8B8,192 contextBest forGeneral ChatCommunity fitCustomQ4_K_MOpen →
COCohereAya Expanse 32BAya Expanse multilingual profile for broad language coverage.Minimum VRAM20 GB20-34 GB comfort rangeModel scale32B131,072 contextBest forGeneral ChatCommunity fitCustomQ4_K_MOpen →
BLBloomBLOOM 3BBigScience's multilingual modelMinimum VRAM2.5 GB2.5-4.5 GB comfort rangeModel scale3B2,048 contextBest forGeneral ChatCommunity fitCustomQ4_K_MOpen →
BLBloomBLOOM 7B1Mid-size BLOOM modelMinimum VRAM5 GB5-9 GB comfort rangeModel scale7.1B2,048 contextBest forGeneral ChatCommunity fitCustomQ4_K_MOpen →
MSMistralCodestral 22BMistral's specialized coding modelMinimum VRAM14 GB14-24 GB comfort rangeModel scale22.2B32,768 contextBest forCodingCommunity fitCustomQ4_K_MOpen →
COCohereCommand ACohere Command A profile for high-quality assistant interactions.Minimum VRAM67 GB67-117 GB comfort rangeModel scale111B131,072 contextBest forGeneral Chat / Math / ReasoningEstimated fitCustomQ4_K_MOpen →
COCohereCohere Command RCohere's open weights model with RAG optimizationMinimum VRAM17 GB17-26 GB comfort rangeModel scale35B128,000 contextBest forGeneral ChatCommunity fitCustomQ4_K_MOpen →
COCohereCohere Command R+Cohere's larger command modelMinimum VRAM36 GB36-62 GB comfort rangeModel scale104B128,000 contextBest forGeneral ChatEstimated fitCustomQ4_K_MOpen →
DSDeepseekDeepSeek Coder V2 LiteEfficient coding model from DeepSeekMinimum VRAM10 GB10-18 GB comfort rangeModel scale16B128,000 contextBest forCodingCommunity fitCustomQ4_K_MOpen →
DSDeepseekDeepSeek R1 671BFrontier-scale DeepSeek R1 profile for extreme-memory clusters.Minimum VRAM229 GB229-390 GB comfort rangeModel scale671B131,072 contextBest forMath / Reasoning / General ChatEstimated fitCustomQ4_K_MOpen →