Vikasit Mini
Lightweight assistant. Summaries, chat, and basic reasoning.
Overview
Vikasit Mini is a 1.7B dense assistant for summarisation, multi-turn chat, and basic reasoning. A strong step up from Nano while still running on modest laptops and edge devices.
Specifications
- Total parameters
- 1.7B
- Architecture
- Dense transformer
- Layers
- 28
- Attention
- GQA (16 query / 8 KV heads), tied embeddings
- Context window
- 32K native
- Vocabulary
- 151,669
- Modalities
- Text in → text out
- License
- Apache 2.0
Capabilities
- Multi-turn chat and summarisation
- Basic reasoning and instruction following
- Laptop / edge friendly
- Thinking and non-thinking modes
Benchmarks
| Benchmark | Score |
|---|---|
| MMLU-Pro | 36.8 |
| GPQA-Diamond | 40.1 |
| AIME 2025 | 36.8 |
| MATH-500 | 93.4 |
| LiveCodeBench v5 | 33.2 |
| BFCL v3 | 56.6 |
| IFEval | 72.5 |
| HumanEval | N/A |
Instruct numbers from the Qwen3 Technical Report; MMLU-Pro is the base-model figure. Thinking-mode scores shown.
Hardware & deployment
| Precision | Memory |
|---|---|
| bf16 | ~3.4 GB |
| INT4 | ~1 GB |
Quick start
Vikasit Mini is an open-weight model. Self-host it with any OpenAI-compatible inference server and call it with the OpenAI SDK as shown below.
# pip install openai
import os
from openai import OpenAI
client = OpenAI(
base_url="http://localhost:8000/v1",
api_key="sk-local", # self-hosted servers accept any token
)
resp = client.chat.completions.create(
model="vikasit-mini",
messages=[
{"role": "user", "content": "Explain Vikasit Mini in one sentence."}
],
)
print(resp.choices[0].message.content)Limitations
- Reasoning depth below 4B+ models
- Limited long-context retention at 32K
Vikasit Mini FAQ
How much does Vikasit Mini cost?
Vikasit Mini is an open-weight model built on Qwen3-1.7B (Apache 2.0). Self-hosting the weights is free under the Apache 2.0 licence — you pay only for the hardware or cloud GPUs you run it on. Typical deployment fits the memory profiles listed in the hardware section above.
Is Vikasit Mini open weight?
Yes. Vikasit Mini is built on Qwen3-1.7B (Apache 2.0) and distributed under the Apache 2.0 licence, so the weights are openly available for self-hosting, fine-tuning, and commercial use, subject to the upstream licence terms.
How do I run Vikasit Mini?
Because Vikasit Mini is open weight, you self-host it with any OpenAI-compatible inference server (such as vLLM or SGLang) loaded with the Qwen3-1.7B (Apache 2.0) weights, then call it with the OpenAI SDK by setting the base URL to your own endpoint.
What context window does Vikasit Mini support?
Vikasit Mini supports a 32K native context window. It is a 1.7B Dense transformer model — full specifications are listed in the table above.
License & attribution
Apache 2.0
Built on Qwen3-1.7B (Apache 2.0). Upstream copyright, license, and attribution notices are retained.