Model Catalogue

Model Catalog

The Model Catalogue is your launchpad for working with large AI models on AI Studio. In a single view, you can discover models, inspect costs and licenses, run live tests in the Playground, and copy ready-to-use API snippets—all before writing a line of production code.

---

What You Can Do in the Catalogue

• Search & filter models by model name, modality (text-generation, multimodal, speech-to-text, and more) or provider.

• Inspect key metadata at a glance: model type, pricing tag, model provider

• Compare alternatives quickly—sort by date of addition and number of parameter.

• Open a Playground session directly from the card to test prompts or upload inputs.

• Copy starter cURL/Python code pre-filled with the correct model string.

• Move to the next step (fine-tune, evaluate, deploy) without leaving the Models section.

---

Supported Modalities

Modality	Typical Use Cases
Text-Generation	Chatbots, summarisation, code completion
Text-to-Embedding	Semantic search, recommendation engines
Image-Text-to-Text	Image captioning, visual Q&A
Speech-to-Text	Transcription, voice-assistant pipelines
Text-to-Speech	Voice synthesis, IVR systems
Multimodal	Combined vision & language reasoning
Tokenizer	Stand-alone tokenisation for offline pipelines

---

Model Card Overview

A Model Card contains three tabs:

Tab	Purpose
Playground	Run interactive tests. Text models expose temperature, top_p, max_tokens, etc.; speech/image models accept uploads.
Starter Code	Copy cURL and Python snippets that call the production endpoint with minimal setup.
Overview	Architecture notes, training data summary, benchmarks, responsible-AI statements, and license details.

Common header tags—pricing, license, GitHub, Base Model / Fine-tuned—let you evaluate suitability at a glance.

---

Quickstart · Trying a Model

1. Locate a model card and open Playground.

2. Enter a prompt or upload an input sample.

3. Adjust parameters if needed and click Run.

4. Review the output.

5. Switch to Starter Code, copy the snippet, and replace "<your secret key here>" with your own token.

---

Choosing the Right Model

Consideration	Practical Guidance
Task Fit	Match the modality to your problem (e.g., speech-to-text for call recordings).
Latency vs Cost	Smaller models are cheaper and faster; larger models often deliver higher quality.
Domain Alignment	Fine-tune when a base model’s answers are too generic for specialised content.
Licensing	Confirm the license permits commercial or derivative use if your product requires it.

---

Requesting Additional Models

If you need a model that is not listed, Please fill the form below with following details:

• Model name & version

• Source link or paper reference

• Intended use case and traffic details

Requests are reviewed weekly and prioritised by feasibility and demand.

---

Current Catalogue Snapshot (July 2025)

Model	Provider	Type
chitrapathak	Krutrim	Image-text-to-text
Llama-3.2-11B-Vision-Instruct	Meta	Image-text-to-text
gemma-3-27b-it	Google	Multimodal
Llama-4-Maverick-17B-128E-Instruct	Meta	Multimodal
Krutrim-Dhwani	Krutrim	Speech-to-Text
DeepSeek-R1-Distill-Llama-70B	deepseek_ai	Text-Generation
DeepSeek-R1-Distill-Llama-8B	deepseek_ai	Text-Generation
DeepSeek-R1	deepseek_ai	Text-Generation
Krutrim-1	Krutrim	Text-Generation
Krutrim-2	Krutrim	Text-Generation
Llama-3.3-70B-Instruct	Meta	Text-Generation
Phi-4-reasoning-plus	Microsoft	Text-Generation
Mistral-7B-v0.2	MistralAI	Text-Generation
Qwen3-30B-A3B	Qwen	Text-Generation
Qwen3-32B	Qwen	Text-Generation
Bhasantarit	Krutrim	Text-to-Embedding
Vyakyarth	Krutrim	Text-to-Embedding
Krutrim-TTS	Krutrim	Text-to-Speech
Krutrim-tokenizer	Krutrim	Tokenizer

For current pricing and rate limits, see the Billing section.

---

Next Steps

• Continue to AI Job → Inferencing to run production jobs.

• Review Billing for token accounting and GPU pricing.

• Consult the API Reference for full endpoint specifications.

---