Models for India

A model for every need.

From compact models that run on a phone to frontier-grade reasoning with million-token context — curated, fine-tuned, and served for Indian languages and real-world workloads.

Text

20 models

vikasit-0.5b-writer

Ultra-light writer. Good for text completion, simple Q&A, and edge devices.

vikasit-writer-0.8b

Improved writer with refined architecture. Mobile and IoT friendly.

Smallest general-purpose model. Autocomplete, quick responses, embedded use.

Lightweight assistant. Summaries, chat, and basic reasoning.

Edge-optimized. Multilingual, 256K context, on-device deployment.

Balanced small model. Good code completion and multi-turn chat.

Next-gen 4B with improved reasoning and multimodal awareness.

Strong mid-range. Solid coding, analysis, and content generation.

vikasit-3-flashPopular

Fast, capable model tuned for high-throughput, low-latency inference.

Strong all-rounder. Complex reasoning, long documents, code review.

27B dense·256K ctx

Powerful dense model. Deep reasoning, advanced coding, research tasks.

vikasit-30b-moe

30B (3B active)·256K ctx

MoE efficiency — 30B quality at 3B inference cost. Fast and smart.

32B dense·256K ctx

Largest dense model on CPU. Best quality for reasoning and code.

vikasit-35b-moe

35B (3B active)·256K ctx

Latest MoE with architecture improvements. Best efficiency/quality ratio.

vikasit-3-coderPopular

80B (3B active)·262K ctx

Code-specialized MoE. FIM support, 262K context, agentic coding.

120B (5B active)·262K ctx

Datacenter MoE. Frontier reasoning at low inference cost — only ~5B active per token.

vikasit-235b-moe

235B (22B active)·262K ctx

Large MoE flagship. Advanced reasoning, agentic workflows, 262K context.

vikasit-reasoner-1t

1T (63B active)·262K ctx

Trillion-scale reasoning MoE. Deep multi-step reasoning, long-horizon agent tasks.

vikasit-titan-1t

1.1T (32B active)·262K ctx

Trillion-parameter agentic MoE. Native multimodal, 262K context, agent-swarm orchestration.

vikasit-titan-1.6tPopular

1.6T (49B active)·1M ctx

Flagship frontier MoE. 1.6T parameters, 1M-token context. Our most capable model.

Vision

3 models

vikasit-vision-2b

Lightweight vision. Image captioning, OCR, visual Q&A on device.

vikasit-vision-4b

Mid-range vision. Document understanding, chart reading, UI analysis.

vikasit-vision-8b

Strong vision. Complex image reasoning, visual code generation.

Voice

3 models

vikasit-voiceSoon

Text-to-speech. Natural voice generation, multilingual support.

vikasit-voice-hdSoon

High-quality TTS. Voice cloning, expressive speech synthesis.

vikasit-omniSoon

30B (3B active)·256K ctx

Full multimodal — text + image + audio in, text + speech out. Real-time.