Models

Vikasit Code hosted models + open-source Vikasit AI model family.

Vikasit Code ships with 5 curated hosted AI models — all available on every subscription plan. You can also run the open-source Vikasit AI models yourself via Ollama or HuggingFace for local or self-hosted inference.

Hosted models

These are served by Vikasit Code’s Pro API (vikasit.ai/zen/v1). Every plan gets access to every model; plans differ only in rate-limit budget.

Model ID	Description	Best for
`vikasit-3-flash`	Snappy, low-latency daily driver	Short chat, simple completions, quick questions
`vikasit-3`	Balanced workhorse — the default	General-purpose coding, refactors, explanations
`vikasit-3-coder`	Coding-specialist with large context	Multi-file refactors, agent loops, debug sessions
`vikasit-3-fast`	Reasoning / thinking-mode model	Architecture decisions, complex problem-solving
`vikasit-3-max`	Flagship — Claude-class quality	Hardest problems, long-horizon planning

The default model is vikasit-3. Most users get the best cost/quality ratio by staying on the default and escalating to vikasit-3-max only when needed.

vikasit-3-flash

Fast, light, and cheap. Ideal for simple tasks where you need a response in under a second.

Context: 128K
Use when: Quick questions, snippet generation, simple completions
Model ID: vikasit-3-flash

vikasit-3

The default daily driver. Strong quality for general coding work, and the most token-budget-friendly option in the lineup.

Context: 128K
Use when: General development, explanations, small-to-medium edits
Model ID: vikasit-3

vikasit-3-coder

Purpose-built for coding agents with a large context window for multi-file understanding.

Context: 256K
Use when: Multi-file refactors, agentic coding, debugging long traces
Model ID: vikasit-3-coder

vikasit-3-fast

Thinking-mode model for tasks that need deliberate reasoning before responding.

Context: 128K
Use when: Architecture decisions, reasoning through complex problems
Model ID: vikasit-3-fast

vikasit-3-max

The flagship model — comparable to Claude Sonnet 4.6 on real-world coding benchmarks.

Context: 128K+
Use when: The hardest problems, mission-critical refactors, quality-above-all tasks
Model ID: vikasit-3-max

Select a model

Once your account is configured, list available models with:

/models

Pick one that matches your task. If unsure, stay on the default vikasit-3.

Set a default

Add the model key to your Vikasit Code config:

{
  "$schema": "https://vikasit.ai/config.json",
  "model": "vikasit-ai/vikasit-3"
}

The format is vikasit-ai/<model-id>.

Small model for background tasks

Vikasit Code uses a separate lightweight model for background jobs like session titles. Configure with small_model:

{
  "$schema": "https://vikasit.ai/config.json",
  "small_model": "vikasit-ai/vikasit-3-flash"
}

By default, Vikasit Code uses vikasit-3-flash for these tasks.

Loading order

When Vikasit Code starts, it picks a model in this priority:

The --model or -m CLI flag

vikasit --model vikasit-ai/vikasit-3-coder

The model key in config
The last-used model
Internal default (vikasit-3)

Plan and access

All 5 hosted models are available on every subscription plan (Lite $10, Pro $50, Max $100). Plans differ only in rate-limit budget. Learn more about plans.

Open-source Vikasit AI models

Want to run inference yourself? The full Vikasit AI model family is open-source and available on Ollama and HuggingFace. 18 models total spanning small-footprint nano variants up to 35B-parameter mixture-of-experts, including code-specialist and vision-capable versions.

General text models

Model	Size	Ollama	HuggingFace
Vikasit Nano	~1B	Vikasit-AI/nano	vikasit-ai/Vikasit-AI-Vikasit-Nano
Vikasit Mini	~1.7B	Vikasit-AI/mini	vikasit-ai/Vikasit-AI-Vikasit-Mini
Vikasit 2B	2B	Vikasit-AI/2b	vikasit-ai/Vikasit-AI-Vikasit-2B
Vikasit 4B	4B	Vikasit-AI/4b	vikasit-ai/Vikasit-AI-Vikasit-4B
Vikasit 3.5 4B	4B	Vikasit-AI/3.5-4b	vikasit-ai/Vikasit-AI-Vikasit-3.5-4B
Vikasit 8B	8B	Vikasit-AI/8b	vikasit-ai/Vikasit-AI-Vikasit-8B
Vikasit 14B	14B	Vikasit-AI/14b	vikasit-ai/Vikasit-AI-Vikasit-14B
Vikasit 27B	27B	Vikasit-AI/27b	vikasit-ai/Vikasit-AI-Vikasit-27B
Vikasit 32B	32B	Vikasit-AI/32b	vikasit-ai/Vikasit-AI-Vikasit-32B
Vikasit 30B MoE	30B-A3	Vikasit-AI/30b-moe	vikasit-ai/Vikasit-AI-Vikasit-30B-MoE
Vikasit 35B MoE	35B-A4	Vikasit-AI/35b-moe	—

Coding-specialist models

Model	Ollama	HuggingFace
Vikasit 3 Flash	Vikasit-AI/3-flash	vikasit-ai/Vikasit-AI-Vikasit-3-Flash
Vikasit 3 Coder	Vikasit-AI/3-coder	vikasit-ai/Vikasit-AI-Vikasit-3-Coder

Vision models

Model	Ollama	HuggingFace
Vikasit Vision 2B	Vikasit-AI/vision-2b	vikasit-ai/Vikasit-AI-Vikasit-Vision-2B
Vikasit Vision 4B	Vikasit-AI/vision-4b	vikasit-ai/Vikasit-AI-Vikasit-Vision-4B
Vikasit Vision 8B	Vikasit-AI/vision-8b	vikasit-ai/Vikasit-AI-Vikasit-Vision-8B

Writer models

Model	HuggingFace
Vikasit 0.5B Writer	vikasit-ai/Vikasit-AI-0.5B-Writer
Vikasit 0.8B Writer	vikasit-ai/Vikasit-AI-0.8B-Writer
Vikasit Writer	Vikasit-AI/writer

Run one locally in 2 minutes

Install Ollama, then pick any model:

ollama pull Vikasit-AI/3-coder
ollama run Vikasit-AI/3-coder

Point Vikasit Code at your local Ollama instance:

{
  "$schema": "https://vikasit.ai/config.json",
  "model": "ollama/Vikasit-AI/3-coder",
  "provider": {
    "ollama": {
      "npm": "@ai-sdk/openai-compatible",
      "api": "http://localhost:11434/v1"
    }
  }
}

All Vikasit AI models are licensed permissively — use them commercially, fine-tune them, self-host them at any scale.