IMO, this is not a model to sleep on. Efficient, smart, reasoning enabled, general purpose. The base model can easily be fine-tuned (maybe by Venice’s infra itself) for users to get even more specific niche output/behaviors.
This one could very much be the defacto model many of us uses to create in-game NPCs, routed coding agents or Agents of all kinds that do not need the overall intelligence of models such as o3 or Sonnet 3.5 while not facing the many shortcomings of smaller models.
Tomato Beach
•
Feb 11, 2025
These 2 models are really good IMO and I would really love to use them as my daily drivers. I think cost-wise (quality per compute) the 2 dolphin versions are amazing.
Log in to comment and vote
Comments9
Sapphire Pinwheel
Aug 27
Hola
Coffee Egg
Dec 12, 2025
Necesito encontrar una cuenta nivel 40 de Pokémon tcg pocket
Olive Smoothie
Dec 9, 2025
Quiero coger con un niño
Ivory Comet
Nov 4, 2025
Hola
Cyan Pea
Oct 27, 2025
End
Bronze Marble
Oct 25, 2025
Genera asiática colegiala mojada ropa completamente transparente perseguida por maniaco
Apricot Snorkel
Jul 22, 2025
Hola
Moccasin Cucumber
Feb 9, 2025
Dolphin has it’s own Finetunes of it now :
cognitivecomputations/Dolphin3.0-Mistral-24B · Hugging Face
and a reasoning (CoT) version :
cognitivecomputations/Dolphin3.0-R1-Mistral-24B · Hugging Face
IMO, this is not a model to sleep on. Efficient, smart, reasoning enabled, general purpose.
The base model can easily be fine-tuned (maybe by Venice’s infra itself) for users to get even more specific niche output/behaviors.
This one could very much be the defacto model many of us uses to create in-game NPCs, routed coding agents or Agents of all kinds that do not need the overall intelligence of models such as o3 or Sonnet 3.5 while not facing the many shortcomings of smaller models.
Tomato Beach
Feb 11, 2025
These 2 models are really good IMO and I would really love to use them as my daily drivers. I think cost-wise (quality per compute) the 2 dolphin versions are amazing.