Dedicated Endpoints

Run this model inference on single tenant GPU with unmatched speed and reliability at scale.

Learn more

Get help setting up a custom Dedicated Endpoints.

Talk with our engineer to get a quote for reserved GPU instances with discounts.

README

Model Details

  • Parameters: 31B
  • Base model: google/gemma-4-31B
  • Focus: multi-turn roleplay, companion chat, regional language handling, code-switching
  • License: Subject to upstream Gemma license terms

Training

Peek was fine-tuned for human-facing Ira companion conversations involving long-running chats, emotional ambiguity, casual friendship dynamics, and roleplay-style continuity. The training set was purpose-built for the Ira companion app.

The training focus includes:

  • sustained multi-turn character consistency
  • natural Hinglish and Indian English conversation
  • regional language code-switching, including Telugu-English, Tamil-English, Malayalam-English, and mixed-register chat
  • emotionally attentive replies without sounding formal or robotic
  • companion-style dialogue that can hold context across extended conversations

Intended Use

Peek is designed for conversational deployments where users expect natural, emotionally aware, multi-turn interaction.

It is intended for:

  • companion-style chat
  • roleplay and character-driven conversation
  • multilingual and code-switched casual conversation
  • emotionally attentive non-critical support
  • long-form conversational engagement

Scope and Ongoing Work

Peek is an Ira-focused release. As with any conversational model, it has scope boundaries we are actively refining:

  • It is not intended as a source of medical, legal, financial, or safety-critical advice.
  • It should not be deployed as a replacement for professional human support.
  • Performance is strongest in conversational, roleplay, and companion-chat scenarios.
  • Regional language quality and code-switching behavior are ongoing areas of improvement.

About

Peek is developed by Rumik for the Ira companion app. For collaboration, deployment questions, or failure-case reports, contact pranav@rumik.ai.

Model provider

rumik-ai

Model tree

Base

google/gemma-4-31B

Fine-tuned

this model

Modalities

Input

Text, Image

Output

Text

Pricing

Dedicated Endpoints

View details

Supported Functionality

Model APIs

Dedicated Endpoints

Container

More information

Explore FriendliAI today