🤖 Sam-reason-S1

Sam-reason-S1 is the first model in the new Reason-S series from SmilyAI, following the experimental V-series (e.g. Sam-reason-v1, v2). This line of models focuses on embedding reasoning, identity, and personality into open-domain conversational agents.


📌 Model Series Context

This model is part of SmilyAI's Reason-S1 series — a continuation of the V-series that aims to produce characterful, logically grounded models. Sam-reason-S1 is designed to:

  • Maintain a consistent AI identity: "Sam by SmilyAI"
  • Use clear reasoning steps formatted like:
    [reason]Here's how I think about it...[/reason]
    Final answer.
    
  • Simulate custom personas, including villain-style or sarcastic tones
  • Respond reliably in open-ended conversations

🛠️ Intended Use

Sam-reason-S1 is intended for:

  • Chatbots with reasoning support
  • Personality-rich AI agents
  • Story characters with logic-driven replies
  • Educational or experimental reasoning assistants

🚀 How to Use

from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline

tokenizer = AutoTokenizer.from_pretrained("Smilyai-labs/Sam-reason-S1")
model = AutoModelForCausalLM.from_pretrained("Smilyai-labs/Sam-reason-S1")

chat = pipeline("text-generation", model=model, tokenizer=tokenizer, max_new_tokens=200)

prompt = "[reason]Thinking...[/reason]
User: What's the capital of France?
Sam:"
response = chat(prompt, do_sample=True, temperature=0.7)[0]["generated_text"]

📁 Files Included

  • model.safetensors
  • config.json
  • tokenizer_config.json
  • Tokenizer files (e.g. tokenizer.model, or vocab.json + merges.txt)
  • generation_config.json (optional)

🧪 Limitations

  • Replies are sampled (not deterministic)
  • Identity and tone may vary depending on prompt phrasing
  • Model has no memory across turns (stateless)
  • Can sometimes confuse itself if the users prompt is very open ended

🧠 Future

  • Sam-reason-S2: Memory-enabled reasoning agent
  • Sam-coder-S1: Code-focused reasoning assistant

👤 Author

Created by SmilyAI
Lead: @boningcui
Repository: Smilyai-labs/Sam-reason-S1
License: MIT


📢 Feedback

Have suggestions or want to contribute ideas to the Reason series? Reach out on Hugging Face or via SmilyAI channels.

Downloads last month
61
Safetensors
Model size
1.35B params
Tensor type
FP16
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for Smilyai-labs/Sam-reason-S1

Quantizations
1 model

Space using Smilyai-labs/Sam-reason-S1 1