Frontier AI, built in India
Vikasit builds open models, a chat assistant, an OpenAI-compatible inference API, and an agentic coding CLI — one intelligence, every surface. Start free with 2M tokens a day.
VS Code
Cursor
Windsurf
VS Codium
Zed
Gitpod
Eclipse TheiaOne AI platform, many surfaces
Vikasit AI is more than a coding tool. Chat, build on our API, code in your terminal, or run our open models — all powered by the same Vikasit intelligence.
Try Vikasit free — no signup
Ask anything. Powered by Vikasit Nova.
Free preview · rate-limited · powered by Vikasit NovaGet unlimited free →
What is Vikasit AI?
Vikasit AI is India's frontier AI company. We build the Vikasit 3 model family and the products on top of it: Vikasit Chat (a multimodal assistant), an OpenAI-compatible inference API, open-weight models you can self-host, and Vikasit Code — an agentic coding CLI. One intelligence, every surface — start free with 2M tokens a day.
Everything you need to code faster
A complete AI development environment that integrates with your existing workflow. No compromises.
Custom AI Agents
9 pre-built agents plus unlimited custom agents. Build specialized assistants for your exact workflow — from code review to planning to exploration.
41+ Built-in Tools
File operations, code analysis, web search, task management, git operations, LSP, notebook editing, batch operations — all available to your AI.
Terminal & Desktop
Native CLI with beautiful TUI, plus Desktop apps for macOS, Windows, and Linux built with Tauri. Deep linking, notifications, and system integration.
MCP & LSP Support
Extend capabilities with Model Context Protocol servers and Language Server Protocol for deep code intelligence across any language.
GitHub Integration
Native PR reviews, issue commenting, workflow automation, GitHub Actions integration with OIDC support — all from your terminal.
Powered by Vikasit AI
Our own AI models from vikasit-3-small to vikasit-3-max, plus Writer models on Ollama & Hugging Face. Also supports 109 third-party providers.
Available Everywhere
One tool, every platform. Works on macOS, Windows, and Linux.
CLI
Terminal-native experience with beautiful TUI. Install via curl, npm, or brew and start coding in seconds.
Desktop App
Native Tauri application with system notifications, deep linking, and auto-updates for a seamless experience.
Web Console
Full-featured dashboard for workspace management, API keys, usage analytics, and team collaboration.
See it in action
Watch Vikasit Code fix a bug in seconds — reading files, applying edits, and running tests automatically.
Powered by Vikasit AI
Purpose-built models for coding, reasoning, and vision — from quick completions to deep thinking.
vikasit-3-small
FastQuick responses for simple tasks
vikasit-3-flash
FlashUltra-fast with streaming output
vikasit-3-coder
CoderOptimized for code generation and refactoring
vikasit-3-fast
BalancedSpeed and quality combined
vikasit-3-vision
VisionImage understanding + code generation
vikasit-3
FlagshipOur most capable general model
vikasit-3-thinking
ReasoningExtended reasoning for complex tasks
vikasit-3-max
MaxMaximum capability, no limits
Which Model Should You Use?
Each model has different capabilities, token costs, and ideal use cases. Here's a practical guide to choosing the right one.
Token Consumption Multiplier
Each model consumes tokens at different rates against your plan's rate limit. Lighter models let you make many more requests with the same quota compared to heavier models.
- •Quick code completions & suggestions
- •Simple bug fixes and small edits
- •Chat-style Q&A about your codebase
- •Boilerplate and scaffold generation
- •Daily coding tasks on a budget
- •Complex multi-file refactoring
- •Deep reasoning or architectural planning
Pro Tip: Use Flash as your daily driver for routine tasks. It has the lowest token consumption of any model, meaning you can make far more requests than heavier models with the same quota.
- •Code generation and refactoring
- •Writing tests and test suites
- •Debugging complex logic errors
- •Code review and optimization
- •Multi-file changes with reasoning
- •Simple one-liner fixes (use Flash instead)
- •Non-coding tasks like documentation
Pro Tip: The sweet spot for most professional developers. Reasoning capability means it plans before editing, catching issues Flash might miss. Moderate consumption is a fair trade for the quality boost.
- •Large-scale refactoring across many files
- •Complex architectural decisions
- •When you need fast responses AND high quality
- •Time-critical production hotfixes
- •Reviewing large PRs with deep context
- •Budget-conscious daily use
- •Simple tasks (massive overkill)
Pro Tip: This is the 397B parameter powerhouse with the highest token consumption of any model. Use it strategically for complex tasks where speed and quality both matter. Don't use it for simple edits.
- •General-purpose development — the ideal daily driver
- •Balanced quality + cost for all coding tasks
- •Understanding large codebases (128K context)
- •Both simple and moderately complex tasks
- •When you want reasoning without high token cost
- •Tasks requiring absolute maximum capability
Pro Tip: The best value model on Max plan. Low token consumption with reasoning enabled and 235B parameters makes this the recommended default for Max subscribers. Set it as your default model.
- •The most complex coding challenges
- •System architecture design and planning
- •Migrating entire codebases between frameworks
- •Solving problems other models can't crack
- •When nothing else gives you the right answer
- •Routine daily tasks (too expensive)
- •Simple edits or quick fixes
Pro Tip: Our largest model at 480B parameters. Reserve it for when vikasit-3 or vikasit-3-coder aren't cutting it. Heavy token consumption means a single complex task can use significant quota. Use wisely.
Our Recommendation
For Lite subscribers: Use vikasit-3-flash as your everyday model. It's fast with minimal token consumption, and handles most routine coding tasks well.
For Pro subscribers: Set vikasit-3-coder as your default. It's purpose-built for coding with reasoning at moderate consumption. Use Flash for quick tasks to conserve quota.
For Max subscribers: Use vikasit-3 (the flagship) as your default — best value with low consumption, 235B parameters and reasoning. Escalate to vikasit-3-max (480B) only for the toughest problems.
Open Source Models
Available on Ollama & Hugging Face. Runs on CPU — no GPU required.
Writer 0.5B
Lightweight text generation model optimized for Indian markets. Runs on CPU — no GPU required.
Writer 0.8B
Higher-capacity writer model for blogs, documentation, and content generation. CPU-first deployment.
Already on OpenAI? Switch in 30 seconds.
The Vikasit Inference API is OpenAI-compatible. Change two lines — the base URL and your key — and keep all your existing code, SDKs, and tools.
- 313M+ tokens served
- 109 providers
- 2M free tokens / day
from openai import OpenAI
client = OpenAI(
base_url="https://api.vikasit.ai/v1", # ← change this
api_key="sk-vikasit-...", # ← and this
)
resp = client.chat.completions.create(
model="vikasit-nova", # free: 2M tokens/day
messages=[{"role": "user", "content": "Hello!"}],
)Simple, transparent pricing
Two ways to build: the Vikasit Code CLI (subscription or pay-as-you-go) and the Vikasit Inference API (pay-as-you-go, with 2M free tokens a day).
Pro — CLI Subscription
Monthly plans with daily token quotas — up to 6 crore tokens a day on Max, enough for 12+ hours of agentic coding. Choose Lite, Pro, or Max based on how much you build.
Go — CLI Pay As You Go
Pay only for what you use with per-token billing. Access all models with no monthly commitment, no minimum spend, and transparent usage tracking.
Inference — API
OpenAI-compatible API for your own apps. Vikasit models plus the whole catalog (DeepSeek, Qwen, Llama, Claude…), pay-as-you-go. Free tier: 2M tokens/day on Vikasit Nova.
Frequently asked questions
Everything you need to know about Vikasit Code.
Vikasit AI is India's frontier AI company. We build the Vikasit 3 model family and the products on top of it: Vikasit Chat (a multimodal assistant), an OpenAI-compatible inference API, open-weight models you can self-host, and Vikasit Code — an agentic coding CLI. One intelligence, every surface — start free with 2M tokens a day.
Yes. Vikasit Code is open source, built on top of OpenCode and released under an MIT/open license. You can read the source, self-host it, and contribute back.
Yes. Vikasit Code is a drop-in, terminal-native alternative to Claude Code, Cursor, and Aider. It's multi-model, supports bring-your-own-key (BYOK) across 109 providers, and exposes an OpenAI-compatible API.
Vikasit Code offers a Pro subscription starting at $10/month (Lite), with Pro at $50/month and Max at $100/month. Prefer pay-as-you-go? Vikasit Go is per-token billing with no monthly commitment. You can also bring your own API keys across 109 providers.
Install via npm with npm i -g @vikasitai/vikasit-code, via the install script with curl -fsSL https://vikasit.ai/install | sh, or via Homebrew with brew install vikasit-ai/tap/vikasit-code. For VS Code, Cursor, and other editors, install the Vikasit Code extension.
Vikasit Code runs on the Vikasit AI model family (from vikasit-3-small to vikasit-3-max, plus vikasit-3-thinking for reasoning) and supports 109 third-party providers — including OpenAI, Anthropic, Google, Mistral, and more — via bring-your-own-key (BYOK).
Yes. Vikasit Code has first-class support for the Model Context Protocol (MCP) servers and Language Server Protocol (LSP) integrations for code intelligence.
Ready to code with AI?
Install Vikasit Code in seconds and supercharge your development workflow with the power of Vikasit AI.