^jJdddddddddd d d d dd ddddd dddddddddd ddddddddd d d!d"d#ddd$dd%d&d d'd(d)dd*d+dd,d-d d.d/d0dd1d2dd3d4d d5d6d7dd8d9dd:d;d gZedlZeejd?eDd@AedBeed>Sd>S)CzQwen3.6-35B-A3B-UDz!Qwen3.6 35B Uncensored (128K ctx)zFQwen3.6 35B A3B uncensored distill with tuned 100K+ long-context flagslocalzQwen3.6-35B-A3B-UD-Q4_K_S.ggufiiz llama-server35b) namedescdetailprovider model_nameportcontextserviceshortz Qwen3.5-9B-UDz Qwen3.5 9B Uncensored (128K ctx)zCQwen3.5 9B uncensored distill with tuned 100K long-context batchingz!qwen3.5-9b-uncensored-Q4_K_M.ggufizllama-server-9b9bz Gemma-4-26BzGemma 4 26B Instruct (131K ctx)zRGoogle Gemma 4 26B IQ4_XS quantized instruction model with strong 100K context fitzgemma-4-26B-IQ4_XS.ggufiizllama-server-gemma4gemma4zQwen3-Coder-30BzQwen3 Coder 30B A3B (65K ctx)z Coding-focused 30B MoE at Q4_K_Mz(Qwen3-Coder-30B-A3B-Instruct-Q4_K_M.ggufi=#izllama-server-codercoder30zQwen3.6-35B-MTPzQwen3.6 35B A3B MTP (65K ctx)zCMulti-Token Prediction variant, faster generation than standard 35Bi>#zllama-server-mtpmtpzQwen2.5-Coder-14BzQwen2.5 Coder 14B (65K ctx)uCDense coding model at Q4_K_M — requires ngl<99 due to VRAM limitsz&qwen2.5-coder-14b-instruct-q4_k_m.ggufi?#zllama-server-coder14coder14zQwen3-30B-2507zQwen3 30B A3B 2507 (65K ctx)z:Qwen3 30B A3B July 2025 release for general 30B comparisonz'Qwen3-30B-A3B-Instruct-2507-Q4_K_M.ggufi@#zllama-server-25072507zQwen3.5-14B-Opusz(Qwen3.5 14B A3B Opus-Distilled (65K ctx)uEClaude 4.6 Opus reasoning-distilled 14B at MXFP4 — fast prompt evalzBQwen3.5-14B-A3B-Claude-Opus-Reasoning-Distilled-4.6-MXFP4_MOE.ggufiA#zllama-server-opus14opus14__main__Ncg|] }|d S)r).0ms '/home/rurouni/.hermes/scripts/models.py rks000Aai000)indentz Total models: )MODELS__name__jsonprintdumpslenrrrr%s%3Z6!   2W9$  1f/(  "/4@'  "/W6%  $-W>)  !.N?&  #:YZ(  ma F zKKK E*$*00000 ; ; ;<<< E *SS[[ * *+++++r