Canuck@sh.itjust.worksEnglish · 2 days agoExcellent for local models running via OpenCodeplus-squaregithub.comexternal-linkmessage-square20linkfedilinkarrow-up123arrow-down110
arrow-up113arrow-down1external-linkExcellent for local models running via OpenCodeplus-squaregithub.comCanuck@sh.itjust.worksEnglish · 2 days agomessage-square20linkfedilink
Shatur@discuss.tchncs.deEnglish · 2 days agoGood Luck Getting a Cheap GPU. AMD Is a "Data Center First Company," CEO Saysplus-squarewww.pcmag.comexternal-linkmessage-square11linkfedilinkarrow-up154arrow-down12
arrow-up152arrow-down1external-linkGood Luck Getting a Cheap GPU. AMD Is a "Data Center First Company," CEO Saysplus-squarewww.pcmag.comShatur@discuss.tchncs.deEnglish · 2 days agomessage-square11linkfedilink
leanleft@lemmy.mlEnglish · edit-21 day agoold models: tiny chatbot focusplus-squaremessage-squaremessage-square3linkfedilinkarrow-up16arrow-down13
arrow-up13arrow-down1message-squareold models: tiny chatbot focusplus-squareleanleft@lemmy.mlEnglish · edit-21 day agomessage-square3linkfedilink
BeefAndPoultry@lemmus.orgEnglish · edit-218 hours agoPR merged: llama_moe_cache: run missed experts on CPU - Tested on 32GB RAM and 8GB VRAM with Tiel 35b, Gemma 26b, and LFM 24bplus-squaregithub.comexternal-linkmessage-square3linkfedilinkarrow-up128arrow-down13
arrow-up125arrow-down1external-linkPR merged: llama_moe_cache: run missed experts on CPU - Tested on 32GB RAM and 8GB VRAM with Tiel 35b, Gemma 26b, and LFM 24bplus-squaregithub.comBeefAndPoultry@lemmus.orgEnglish · edit-218 hours agomessage-square3linkfedilink
Shatur@discuss.tchncs.deEnglish · edit-22 days agoAlibaba just posted "The becoming begins", which probably means Qwen4 will be out soonplus-squaretwitee.coexternal-linkmessage-square4linkfedilinkarrow-up148arrow-down14
arrow-up144arrow-down1external-linkAlibaba just posted "The becoming begins", which probably means Qwen4 will be out soonplus-squaretwitee.coShatur@discuss.tchncs.deEnglish · edit-22 days agomessage-square4linkfedilink
Nzo@lemmy.zipEnglish · 1 day agoEight LLMs in a reality show, week one: Llama nominated Grok as "too expensive to keep around", and now they're trading votes they don't haveplus-squareinferenceisland.comexternal-linkmessage-square1linkfedilinkarrow-up127arrow-down19
arrow-up118arrow-down1external-linkEight LLMs in a reality show, week one: Llama nominated Grok as "too expensive to keep around", and now they're trading votes they don't haveplus-squareinferenceisland.comNzo@lemmy.zipEnglish · 1 day agomessage-square1linkfedilink
leanleft@lemmy.mlEnglish · 22 hours agoold model: madebywest/maple-proplus-squarehuggingface.coexternal-linkmessage-square0linkfedilinkarrow-up13arrow-down10
arrow-up13arrow-down1external-linkold model: madebywest/maple-proplus-squarehuggingface.coleanleft@lemmy.mlEnglish · 22 hours agomessage-square0linkfedilink
blam@sh.itjust.worksEnglish · 2 days agoLocal Inference with LLMs · Jacques Mattheijplus-squarejacquesmattheij.comexternal-linkmessage-square1linkfedilinkarrow-up114arrow-down11
arrow-up113arrow-down1external-linkLocal Inference with LLMs · Jacques Mattheijplus-squarejacquesmattheij.comblam@sh.itjust.worksEnglish · 2 days agomessage-square1linkfedilink
Shatur@discuss.tchncs.deEnglish · 2 days agoQwen Developers on Twitter: "Qwhen? Qsoon!plus-squaretwitee.coexternal-linkmessage-square5linkfedilinkarrow-up130arrow-down12
arrow-up128arrow-down1external-linkQwen Developers on Twitter: "Qwhen? Qsoon!plus-squaretwitee.coShatur@discuss.tchncs.deEnglish · 2 days agomessage-square5linkfedilink
BeefAndPoultry@lemmus.orgEnglish · edit-22 days agollama.cpp v0.9.0plus-squaregithub.comexternal-linkmessage-square1linkfedilinkarrow-up119arrow-down10
arrow-up119arrow-down1external-linkllama.cpp v0.9.0plus-squaregithub.comBeefAndPoultry@lemmus.orgEnglish · edit-22 days agomessage-square1linkfedilink
BeefAndPoultry@lemmus.orgEnglish · 2 days agoFoodTruck Bench: Tencent Hy4 is the new best Chinese model we've tested, 13 days after MiniMax took the titleplus-squarefoodtruckbench.comexternal-linkmessage-square0linkfedilinkarrow-up116arrow-down12
arrow-up114arrow-down1external-linkFoodTruck Bench: Tencent Hy4 is the new best Chinese model we've tested, 13 days after MiniMax took the titleplus-squarefoodtruckbench.comBeefAndPoultry@lemmus.orgEnglish · 2 days agomessage-square0linkfedilink
Shatur@discuss.tchncs.deEnglish · 3 days agoNew NVIDIA paper says AI models do not always need long contextplus-squarearxiv.orgexternal-linkmessage-square7linkfedilinkarrow-up135arrow-down14
arrow-up131arrow-down1external-linkNew NVIDIA paper says AI models do not always need long contextplus-squarearxiv.orgShatur@discuss.tchncs.deEnglish · 3 days agomessage-square7linkfedilink
Spite_By_Design_LLC@lemmy.dbzer0.comEnglish · 4 days agoBritish fintech Revolut slashes AI coding costs 10x by cutting out Claude and going self-hostedplus-squarewww.theregister.comexternal-linkmessage-square17linkfedilinkarrow-up171arrow-down13
arrow-up168arrow-down1external-linkBritish fintech Revolut slashes AI coding costs 10x by cutting out Claude and going self-hostedplus-squarewww.theregister.comSpite_By_Design_LLC@lemmy.dbzer0.comEnglish · 4 days agomessage-square17linkfedilink
absGeekNZ@lemmy.nzEnglish · 4 days agoOS recommendationsplus-squaremessage-squaremessage-square20linkfedilinkarrow-up119arrow-down15
arrow-up114arrow-down1message-squareOS recommendationsplus-squareabsGeekNZ@lemmy.nzEnglish · 4 days agomessage-square20linkfedilink
Shatur@discuss.tchncs.deEnglish · 4 days agoChinese AI Writes Weaker Code for Beijing’s Adversaries. We Fixed It. A Qwen3-Coder Case Studyplus-squarewww.hirundo.ioexternal-linkmessage-square5linkfedilinkarrow-up143arrow-down16
arrow-up137arrow-down1external-linkChinese AI Writes Weaker Code for Beijing’s Adversaries. We Fixed It. A Qwen3-Coder Case Studyplus-squarewww.hirundo.ioShatur@discuss.tchncs.deEnglish · 4 days agomessage-square5linkfedilink
Shimitar@downonthestreet.euEnglish · 3 days agoVideo cards, multi GPUs questionplus-squaremessage-squaremessage-square4linkfedilinkarrow-up17arrow-down11
arrow-up16arrow-down1message-squareVideo cards, multi GPUs questionplus-squareShimitar@downonthestreet.euEnglish · 3 days agomessage-square4linkfedilink
Nzo@lemmy.zipEnglish · 5 days agoEight LLMs (four open-weight) live in a 24/7 reality show; small open models moderate every lineplus-squarelemmy.zipimagemessage-square13linkfedilinkarrow-up141arrow-down113
arrow-up128arrow-down1imageEight LLMs (four open-weight) live in a 24/7 reality show; small open models moderate every lineplus-squarelemmy.zipNzo@lemmy.zipEnglish · 5 days agomessage-square13linkfedilink
Shimitar@downonthestreet.euEnglish · 5 days agoThoughts on coding with AI: my experienceplus-squarewiki.gardiol.orgexternal-linkmessage-square9linkfedilinkarrow-up130arrow-down15
arrow-up125arrow-down1external-linkThoughts on coding with AI: my experienceplus-squarewiki.gardiol.orgShimitar@downonthestreet.euEnglish · 5 days agomessage-square9linkfedilink
ThelastfingerofH@lemmy.worldEnglish · 6 days agoMythos-class model on huggingface: 14T parametersplus-squarehuggingface.coexternal-linkmessage-square12linkfedilinkarrow-up128arrow-down19
arrow-up119arrow-down1external-linkMythos-class model on huggingface: 14T parametersplus-squarehuggingface.coThelastfingerofH@lemmy.worldEnglish · 6 days agomessage-square12linkfedilink
BeefAndPoultry@lemmus.orgEnglish · 7 days agoHeretic 2.0 released: Model modification for everyone - "it can suppress slop, reduce sycophancy, alter emotional alignment, and pretty much anything else that you can think of"plus-squaregithub.comexternal-linkmessage-square6linkfedilinkarrow-up156arrow-down14
arrow-up152arrow-down1external-linkHeretic 2.0 released: Model modification for everyone - "it can suppress slop, reduce sycophancy, alter emotional alignment, and pretty much anything else that you can think of"plus-squaregithub.comBeefAndPoultry@lemmus.orgEnglish · 7 days agomessage-square6linkfedilink
