[{"data":1,"prerenderedAt":278},["ShallowReactive",2],{"docs-en-/en/docs/service/ai/local":3},{"id":4,"title":5,"body":6,"description":270,"extension":271,"meta":272,"navigation":273,"path":274,"seo":275,"stem":276,"__hash__":277},"docs/en/docs/service/ai/local.md","AI Services (Local)",{"type":7,"value":8,"toc":259},"minimark",[9,13,17,24,29,32,37,40,44,47,51,54,58,62,65,68,72,75,79,137,140,143,151,154,161,167,170,173,176,179,185,191,194,197,200,203,209,215,219,222,244],[10,11,5],"h1",{"id":12},"ai-services-local",[14,15,16],"p",{},"Doco Translate can connect to local AI model runtimes such as Ollama, LM Studio, and oMLX for translation.",[14,18,19],{},[20,21],"img",{"alt":22,"src":23},"Local AI translation services","/images/docs/local-ai-services.webp",[25,26,28],"h2",{"id":27},"advantages-of-local-models","Advantages of Local Models",[14,30,31],{},"Using local models is the approach Doco Translate recommends most strongly because it offers several important advantages.",[33,34,36],"h4",{"id":35},"translation-quality","Translation Quality",[14,38,39],{},"Large language models excel at natural-language tasks, especially translation. They usually provide better translation quality than conventional machine translation engines.",[33,41,43],{"id":42},"privacy","Privacy",[14,45,46],{},"Models loaded through a local runtime keep your data on your device. Nothing is sent to a third-party provider, ensuring that document processing remains private.",[33,48,50],{"id":49},"cost","Cost",[14,52,53],{},"Unlike cloud AI APIs, local models do not charge for tokens and have no additional usage fees.",[25,55,57],{"id":56},"limitations-of-local-models","Limitations of Local Models",[33,59,61],{"id":60},"model-size","Model Size",[14,63,64],{},"Local models are usually open source, and models that run well on consumer hardware often have relatively small parameter counts, such as 4B, 9B, 35B-A3B, or 27B. Their general capabilities may therefore be weaker than large cloud models.",[14,66,67],{},"However, translation does not require a very long reasoning chain or the strongest possible reasoning model. Smaller local models are capable of producing good translations, making them a highly cost-effective choice.",[33,69,71],{"id":70},"inference-speed","Inference Speed",[14,73,74],{},"Local inference speed depends on your hardware and the size of the model. Better hardware and smaller models are faster, although local inference is generally still slower than a cloud AI API.",[25,76,78],{"id":77},"local-ai-model-runtimes","Local AI Model Runtimes",[80,81,82,95],"table",{},[83,84,85],"thead",{},[86,87,88,92],"tr",{},[89,90,91],"th",{},"Tool",[89,93,94],{},"Characteristics",[96,97,98,113,125],"tbody",{},[86,99,100,110],{},[101,102,103],"td",{},[104,105,109],"a",{"href":106,"rel":107},"https://ollama.com/",[108],"nofollow","Ollama",[101,111,112],{},"Simple command-line setup and model management",[86,114,115,122],{},[101,116,117],{},[104,118,121],{"href":119,"rel":120},"https://lmstudio.ai/",[108],"LM Studio",[101,123,124],{},"Friendly graphical interface for downloading and running local models",[86,126,127,134],{},[101,128,129],{},[104,130,133],{"href":131,"rel":132},"https://omlx.ai/",[108],"oMLX",[101,135,136],{},"High-speed local LLM inference optimized for Apple silicon",[25,138,109],{"id":139},"ollama",[14,141,142],{},"Ollama is an approachable command-line tool for running local models.",[14,144,145,146,150],{},"Install Ollama by following its official instructions, then run ",[147,148,149],"code",{},"ollama pull MODEL_NAME"," to download a model.",[14,152,153],{},"After the model has downloaded and is running successfully in Ollama, connect it to Doco Translate with one click.",[14,155,156,157,160],{},"The default Ollama endpoint is ",[147,158,159],{},"http://localhost:11434",".",[14,162,163],{},[20,164],{"alt":165,"src":166},"Ollama service settings","/images/docs/ollama-settings.webp",[25,168,121],{"id":169},"lm-studio",[14,171,172],{},"LM Studio provides a graphical interface for running local models and is easy to get started with.",[14,174,175],{},"Install LM Studio from its official website, then download a model inside the app.",[14,177,178],{},"After the model has downloaded, start LM Studio's local model server so that it exposes an API. You can then connect it to Doco Translate with one click.",[14,180,181,182,160],{},"The default LM Studio endpoint is ",[147,183,184],{},"http://localhost:1234",[14,186,187],{},[20,188],{"alt":189,"src":190},"LM Studio service settings","/images/docs/lm-studio-settings.webp",[25,192,133],{"id":193},"omlx",[14,195,196],{},"oMLX is a local model runtime optimized for Apple silicon and provides very fast inference.",[14,198,199],{},"Install oMLX from its official website, open the oMLX console, and download a model.",[14,201,202],{},"oMLX starts its local service automatically. After the model has downloaded and passed a test, connect it to Doco Translate with one click.",[14,204,205,206,160],{},"The default oMLX endpoint is ",[147,207,208],{},"http://localhost:8000",[14,210,211],{},[20,212],{"alt":213,"src":214},"oMLX service settings","/images/docs/omlx-settings.webp",[25,216,218],{"id":217},"recommended-models","Recommended Models",[14,220,221],{},"Doco Translate recommends the following models for translation:",[223,224,225,229,232,235,238,241],"ul",{},[226,227,228],"li",{},"Hy-MT2-1.8B: very fast, minimal memory use, and optimized specifically for translation",[226,230,231],{},"Qwen3.5:9B: fast with relatively low memory use",[226,233,234],{},"Qwen3.6:35B-A3B: strong capabilities and fast performance, with higher memory use",[226,236,237],{},"Gemma4:12B: a balanced choice for quality, speed, and memory use",[226,239,240],{},"Qwen3.6:27B: the strongest capabilities, but slower and memory intensive",[226,242,243],{},"Qwen3.5:4B: fast with low memory use, but slightly less stable",[14,245,246,247,250,251,254,255,258],{},"For most workloads, use ",[147,248,249],{},"Hy-MT2-1.8B"," or ",[147,252,253],{},"Qwen3.5:9B"," as the primary model. Both are fast and memory efficient. For higher quality and stability, use ",[147,256,257],{},"Qwen3.6:35B-A3B",", which is more capable but consumes more memory.",{"title":260,"searchDepth":261,"depth":261,"links":262},"",2,[263,264,265,266,267,268,269],{"id":27,"depth":261,"text":28},{"id":56,"depth":261,"text":57},{"id":77,"depth":261,"text":78},{"id":139,"depth":261,"text":109},{"id":169,"depth":261,"text":121},{"id":193,"depth":261,"text":133},{"id":217,"depth":261,"text":218},"Connect Ollama, LM Studio, or oMLX and translate with AI models running on your device.","md",{},true,"/en/docs/service/ai/local",{"title":5,"description":270},"en/docs/service/ai/local","B9ae3HCPkidNweDAa7Sf2GfIZ4yRRF0VaAXRm3HQOpk",1787207634217]