Experimental reasoning focus
According to the model card, this experimental model is instruction-tuned for nuanced reasoning, dialogue management, retrieval, and summarization.
Open Source Model Profile · EpistemeAI
ReasoningCore-3B-0
ReasoningCore-3B-0 is a 3.21B-parameter Llama reasoning fine-tune from EpistemeAI. According to the model card, it is experimental and targets reasoning, dialogue, retrieval, and summarization.
- Publisher
- EpistemeAI
- Task
- text-generation
- Model type
- llama
- License
- llama3.2
- Library
- transformers
- Publication status
- Accepted · not indexed
Model overview
ReasoningCore-3B-0 is published by EpistemeAI as a text-generation model. The captured configuration identifies LlamaForCausalLM with model type llama, and Safetensors metadata reports 3,212,749,824 parameters. According to the model card, it is an experimental reasoning-enhanced model tuned for dialogue, retrieval, and summarization, with GRPO, supervised learning, and RLHF alignment. Hub tags record unsloth, TRL, conversational, and gsm8k dataset markers.
Recorded capabilities
GRPO plus RLHF
According to the model card, alignment used Group Robust Preference Optimization with supervised learning and reinforcement learning from human feedback.
128k multilingual claims
According to the model card, the model reports a 128k context with multilingual text and code support across eight officially listed languages.
Structured reasoning prompt
According to the model card, generation uses a reasoning-tagged system prompt with bfloat16 transformers loading and boxed math answers.
Use cases in the source record
- Reasoning, dialogue, retrieval, and summarization workflows that follow the card's structured reasoning system prompt.
- Step-by-step math workflows that use the card's boxed-answer system instruction with bfloat16 transformers loading.
Limitations and unknowns
- License needs confirmation: the record field is llama3.2 and the card cites the Llama 3.2 Community License, while the card footer separately mentions apache-2.0.
- Leaderboard-labeled figures appear in captured page text, but no structured evaluation results were extracted; they are not presented as verified scores.
- According to the model card, this is experimental and a newer ReasoningCore-3B-RE1-V2 is available.
- Provider state is historical snapshot data and should be refreshed before being presented as current.
Source and provenance
Source: EpistemeAI/ReasoningCore-3B-0
Captured: Unknown. Processed: 2026-09-07T19:35:05.235350+00:00.
A better version is available: ReasoningCore-3B-RE1-V2 Note: This is an experimental model. ReasomingCore‑3B ReasomingCore‑3B is a multilingual, reasoning‑enhanced large language model developed by EpitemeAI. Pretrained on vast amounts of publicly available data and instruction‑tuned to excel at nuanced reasoning, dialogue management, retrieval, and summarization tasks, it often outperforms many current open source and proprietary conversational models on a range of industry benchmarks. Model Information Model Developer: EpitemeAI Model Architecture: ReasomingCore‑3B is an auto‑regressive language model built on an optimized transfo…
F001F002F003F004F005F006F007F008F010F011F012F013F015F017F018F019F022F023F024