Filter by type:
Filter by tags:
arcee-ai_afm-4.5b
AFM-4.5B is a 4.5 billion parameter instruction-tuned model developed by Arcee.ai, designed for enterprise-grade performance across diverse deployment environments from cloud to edge. The base model was trained on a dataset of 8 trillion tokens, comprising 6.5 trillion tokens of general pretraining data followed by 1.5 trillion tokens of midtraining data with enhanced focus on mathematical reasoning and code generation. Following pretraining, the model underwent supervised fine-tuning on high-quality instruction datasets. The instruction-tuned model was further refined through reinforcement learning on verifiable rewards as well as for human preference. We use a modified version of TorchTitan for pretraining, Axolotl for supervised fine-tuning, and a modified version of Verifiers for reinforcement learning.
The development of AFM-4.5B prioritized data quality as a fundamental requirement for achieving robust model performance. We collaborated with DatologyAI, a company specializing in large-scale data curation. DatologyAI's curation pipeline integrates a suite of proprietary algorithms—model-based quality filtering, embedding-based curation, target distribution-matching, source mixing, and synthetic data. Their expertise enabled the creation of a curated dataset tailored to support strong real-world performance.
The model architecture follows a standard transformer decoder-only design based on Vaswani et al., incorporating several key modifications for enhanced performance and efficiency. Notable architectural features include grouped query attention for improved inference efficiency and ReLU^2 activation functions instead of SwiGLU to enable sparsification while maintaining or exceeding performance benchmarks.
The model available in this repo is the instruct model following supervised fine-tuning and reinforcement learning.
Repository: localaiLicense: apache-2.0
arcee-ai_homunculus
Homunculus is a 12 billion-parameter instruction model distilled from Qwen3-235B onto the Mistral-Nemo backbone. It was purpose-built to preserve Qwen’s two-mode interaction style—/think (deliberate chain-of-thought) and /nothink (concise answers)—while running on a single consumer GPU.
Repository: localaiLicense: apache-2.0
suayptalha_maestro-10b
Maestro-10B is a 10 billion parameter model fine-tuned from Virtuoso-Lite, a next-generation language model developed by arcee-ai. Virtuoso-Lite itself is based on the Llama-3 architecture, distilled from Deepseek-v3 using approximately 1.1 billion tokens/logits. This distillation process allows Virtuoso-Lite to achieve robust performance with a smaller parameter count, excelling in reasoning, code generation, and mathematical problem-solving. Maestro-10B inherits these strengths from its base model, Virtuoso-Lite, and further enhances them through fine-tuning on the OpenOrca dataset. This combination of a distilled base model and targeted fine-tuning makes Maestro-10B a powerful and efficient language model.
Repository: localaiLicense: falcon-llm-license
llama-3.1-supernova-lite-reflection-v1.0-i1
This model is a LoRA adaptation of arcee-ai/Llama-3.1-SuperNova-Lite on thesven/Reflective-MAGLLAMA-v0.1.1. This has been a simple experiment into reflection and the model appears to perform adequately, though I am unsure if it is a large improvement.
Repository: localaiLicense: llama3.1
llama-3.1-supernova-lite
Llama-3.1-SuperNova-Lite is an 8B parameter model developed by Arcee.ai, based on the Llama-3.1-8B-Instruct architecture. It is a distilled version of the larger Llama-3.1-405B-Instruct model, leveraging offline logits extracted from the 405B parameter variant. This 8B variation of Llama-3.1-SuperNova maintains high performance while offering exceptional instruction-following capabilities and domain-specific adaptability.
The model was trained using a state-of-the-art distillation pipeline and an instruction dataset generated with EvolKit, ensuring accuracy and efficiency across a wide range of tasks. For more information on its training, visit blog.arcee.ai.
Llama-3.1-SuperNova-Lite excels in both benchmark performance and real-world applications, providing the power of large-scale models in a more compact, efficient form ideal for organizations seeking high performance with reduced resource requirements.
Repository: localaiLicense: llama3
baldur-8b
An finetune of the L3.1 instruct distill done by Arcee, The intent of this model is to have differing prose then my other releases, in my testing it has achieved this and avoiding using common -isms frequently and has a differing flavor then my other models.
Repository: localaiLicense: agpl-3.0
tulu-3.1-8b-supernova-i1
The following models were included in the merge:
meditsolutions/Llama-3.1-MedIT-SUN-8B
allenai/Llama-3.1-Tulu-3-8B
arcee-ai/Llama-3.1-SuperNova-Lite
Repository: localaiLicense: llama3.1
l3.1-purosani-2-8b
The following models were included in the merge:
hf-100/Llama-3-Spellbound-Instruct-8B-0.3
arcee-ai/Llama-3.1-SuperNova-Lite + grimjim/Llama-3-Instruct-abliteration-LoRA-8B
THUDM/LongWriter-llama3.1-8b + ResplendentAI/Smarts_Llama3
djuna/L3.1-Suze-Vume-2-calc
djuna/L3.1-ForStHS + Blackroot/Llama-3-8B-Abomination-LORA
Repository: localaiLicense: llama3.1
arcee-ai_arcee-maestro-7b-preview
Arcee-Maestro-7B-Preview (7B) is Arcee's first reasoning model trained with reinforment learning. It is based on the Qwen2.5-7B DeepSeek-R1 distillation DeepSeek-R1-Distill-Qwen-7B with further GRPO training. Though this is just a preview of our upcoming work, it already shows promising improvements to mathematical and coding abilities across a range of tasks.
Repository: localaiLicense: apache-2.0