Big (100B+) LLMs (OUTDATED)
Collection of various big LLMs(Not MOE).
Text Generation • Updated • 1.89k • 223Note Merge of xwin and euryale 1.3. The OG that started the big frankenmerging. Creative and smart.
alpindale/miquella-120b
Text Generation • Updated • 11 • 20Note Merge of Euryale and Miqu Q5_K_M. Brain damaged(tested only old version).
grimulkan/Goliath-longLORA-120b-rope8-32k-fp16
Text Generation • Updated • 15 • 7Note 32k context version of Goliath. Dumber and less coherent than the OG Goliath. Has trouble with formatting on short context, has partially lost Alpaca. Confirmed that it works at 17k, but loses minor details of the story, just like the other models extended with this method.
migtissera/Tess-XL-v1.0
Text Generation • Updated • 1.58k • 13Note A finetune of Goliath.
ChuckMcSneed/Dicephal-123B
Text Generation • Updated • 11 • 4Note Self-merge of llama 2 70b. More creative and smarter than the base, but just as disobedient. Should be good for story writing.
ChuckMcSneed/DoubleGold-v0.1-123b-32k
Text Generation • Updated • 8 • 1Note First 32k context 120b frankenstein. Self-merge of grimulkan/aurelian-alpha0.1-70b-rope8-32K-fp16. Dumber and less coherent than Goliath, but compensates for it with large context. Stays coherent at 17k, not sure how well it works at 32k.
ChuckMcSneed/DoubleGold-v0.5-123b-32k
Text Generation • Updated • 9Note Self-merge of grimulkan/aurelian-v0.5-70b-rope8-32K-fp16. Like DoubleGold 0.1, but with much less GPTism.
ChuckMcSneed/WinterGoliath-123b
Text Generation • Updated • 67 • 7Note Merge of Xwin and WinterGoddess. Smarter than Goliath. Likely a bit more aligned.
ChuckMcSneed/WinterGoliath-123b-32k
Text Generation • Updated • 7Note 32k version of WinterGoliath. Comes with all the problems that the extended context creates. Still smart.
ChuckMcSneed/BenchmaxxxerPS-v1-123b
Text Generation • Updated • 9 • 1Note Meme merge meant to beat the benchmark. Has two personalities: angry hobo and an SJW. Not really worth using, unless you want to talk to a digital schizo.
sophosympatheia/Aurora-Nights-103B-v1.0
Text Generation • Updated • 10 • 14Note Self-merge of Aurora-Nights-70B-v1.0. Decent model, a bit more "fun" version of Aurora-Nights.
sophosympatheia/Rogue-Rose-103b-v0.2
Text Generation • Updated • 438 • 21Note This model is a frankenmerge of two custom 70b merges descended from xwin-stellarbright-erp-70b-v2 model.
sophosympatheia/Midnight-Rose-103B-v2.0.3
Text Generation • Updated • 52 • 7Note Self-merge of sophosympatheia/Midnight-Rose-70B-v2.0.3.
nsfwthrowitaway69/Venus-120b-v1.0
Text Generation • Updated • 13 • 27Note Merge of Euryale-1.3-L2-70B, Nous-Hermes-Llama2-70b, and SynthIA-70B-v1.5. Drier than Goliath.
nsfwthrowitaway69/Venus-103b-v1.0
Text Generation • Updated • 6 • 15Note A smaller merge of Euryale-1.3-L2-70B, Nous-Hermes-Llama2-70b, and SynthIA-70B-v1.5.
nsfwthrowitaway69/Venus-120b-v1.1
Text Generation • Updated • 6 • 7Note A result of interleaving layers of Sao10K/Euryale-1.3-L2-70B, Xwin-LM/Xwin-LM-70B-V0.1, and migtissera/SynthIA-70B-v1.5.
nsfwthrowitaway69/Venus-103b-v1.1
Text Generation • Updated • 3 • 11Note A smaller merge of Sao10K/Euryale-1.3-L2-70B, Xwin-LM/Xwin-LM-70B-V0.1, and migtissera/SynthIA-70B-v1.5.
nsfwthrowitaway69/Venus-103b-v1.2
Text Generation • Updated • 3 • 6Note A result of interleaving layers of Sao10K/Euryale-1.3-L2-70B and GOAT-AI/GOAT-70B-Storytelling.
nsfwthrowitaway69/Venus-120b-v1.2
Text Generation • Updated • 300 • 55Note Self-merge of lzlv. Are you going to call all your models "Venus" regardless of what you used to merge? Your naming scheme is leading to confusion.
cognitivecomputations/MegaDolphin-120b
Text Generation • Updated • 984 • 69Note Self-merge of Dolphin-2.2-70b. Very fucking aggresive(GOOD!). Dumber than Goliath.
cognitivecomputations/Samantha-120b
Text Generation • Updated • 14 • 39Note Self-merge of Samantha-1.11-70b.
MarsupialAI/KitchenSink_103b
Text Generation • Updated • 26 • 9Note Merge of royallab/Aetheria-L2-70B, lizpreciatior/lzlv_70b_fp16_hf, Sao10K/WinterGoddess-1.4x-70B-L2.
ibivibiv/chimera-120b
Text Generation • Updated • 13 • 1Note This model is a sliced passthrough merge of Sheep Duck, Xwin, and Sythia.
ibivibiv/giant-macaroni-120b
Text Generation • Updated • 12 • 1Note Merge of AIDC-ai-business/Marcoroni-70B-v1 garage-bAInd/Platypus2-70B-instruct fangloveskari/ORCA_LLaMA_70B_QLoRA.
ibivibiv/athena-120b
Text Generation • Updated • 14 • 1Note Merge of Sao10K/WinterGoddess-1.4x-70B-L2, Xwin-LM/Xwin-LM-70B-V0.1, and migtissera/SynthIA-70B-v1.5.
llmixer/Xwinter-120b
Text Generation • Updated • 11 • 1Note A Goliath-120b style frankenmerge of Xwin-LM-70b-v0.1 and WinterGoddess-1.4x-70b. Similar to ChuckMcSneed/WinterGoliath-123b.
llmixer/BigLiz-120b
Text Generation • Updated • 10 • 1Note A Goliath-120b style frankenmerge of lzlv-70b and WinterGoddess-1.4x-70b.
llmixer/BigAurelian-v0.5-120b-32k
Text Generation • Updated • 15 • 1Note A Goliath-120b style frankenmerge of aurelian-v0.5-70b-32K and WinterGoddess-1.4x-70b. Not sure if it is really 32k.
llmixer/BigWeave-v15-103b
Text Generation • Updated • 59 • 2Note ???
MatrixC7/Meidebenne-120b-v1.0
Text Generation • Updated • 5Note This is a self-merge of sophosympatheia/Midnight-Rose-70B-v1.0.
wolfram/miqu-1-120b
Text Generation • Updated • 59 • 49Note A self-merge of Miqu. Quite good, but has mistralisms. Best 32k model I've tested so far.
wolfram/miquliz-120b
Text Generation • Updated • 6 • 7Note This is a 120b frankenmerge created by interleaving layers of miqu-1-70b-sf with lzlv_70b_fp16_hf using mergekit.
DiscoResearch/DiscoLM-120b
Text Generation • Updated • 1.57k • 22Note A finetune of Goliath.
ludis/tsukasa-120b-qlora
Text Generation • Updated • 3 • 4Note A finetune of Goliath.
ewof/koishi-120b-qlora-gptq
Text Generation • Updated • 3 • 1Note A finetune of Goliath.
dataautogpt3/miqu-120b
Text Generation • Updated • 10 • 5Note Self-merge of Miqu. Actually 137B.
Dracones/perky-103b-v0.1
Text Generation • Updated • 6 • 1Note Self-merge of Dracones/perky-70b-v0.1.
abacusai/TheProfessor-155b
Text Generation • Updated • 92 • 92Note Merge of cognitivecomputations/dolphin-2.2-70b, WizardLM/WizardMath-70B-V1.0, migtissera/SynthIA-70B-v1.2b, epfl-llm/meditron-70b.
deepnight-research/saily_100b
Text Generation • Updated • 1.02k • 7Note Frankenmerged model based on Llama2-chat. Actually 118B. Not as bad as Saily 220, but still kinda bad. Gets easily outperformed by some 70Bs.
deepnight-research/Saily_220B
Text Generation • Updated • 1.59k • 20Note Actually 208b. A frankenmerge of 10(!) different llama2 finetunes. Not worth using. Has all the problems(fake words, incoherence) of frankenmerges without any of the benefits. Can't even follow the prompt most of the time.
tiiuae/falcon-180B
Text Generation • Updated • 5.42k • 1.12kNote Good model, but has a critical flaw: 2k context.
tiiuae/falcon-180B-chat
Text Generation • Updated • 1.03k • 539Note Good model, but has a critical flaw: 2k context.
facebook/galactica-120b
Text Generation • Updated • 793 • 151Note Facebooks failed science model. I can't run this one in llama.cpp lol.
yandex/yalm-100b
Updated • 128Note LLM made by russian company yandex. Can't run it because no gguf.
BlackSamorez/HuYaLM-100B-fp16
Updated • 3 • 16Note Transformers-compatible implementation of the YaLM 100B model. This model has a funny name, if you know russian. Can't run it because no gguf.