Vikasit 32B
Largest dense model. Best quality for reasoning and code.
Overview
Vikasit 32B is the largest dense model in the lineup — top dense-model quality for reasoning, coding, and research. Best when you want maximum quality from a non-MoE architecture.
Specifications
- Total parameters
- 32.8B
- Architecture
- Dense transformer
- Layers
- 64
- Attention
- GQA (64 query / 8 KV heads)
- Context window
- 32K native, 131K via YaRN
- Vocabulary
- 151,669
- Modalities
- Text in → text out
- License
- Apache 2.0
Capabilities
- Best dense-model reasoning and coding
- Research-grade analysis
- 131K extended context (YaRN)
- Thinking and non-thinking modes
Benchmarks
| Benchmark | Score |
|---|---|
| MMLU-Pro | 65.5 |
| GPQA-Diamond | 68.4 |
| AIME 2025 | 72.9 |
| MATH-500 | 97.2 |
| LiveCodeBench v5 | 65.7 |
| BFCL v3 | 70.3 |
| IFEval | 85.0 |
| 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 | ~66 GB |
| INT4 | ~20 GB |
Quick start
Vikasit 32B 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-32b",
messages=[
{"role": "user", "content": "Explain Vikasit 32B in one sentence."}
],
)
print(resp.choices[0].message.content)Limitations
- Higher serving cost than equivalent MoE
- Latency higher than sparse models at scale
Vikasit 32B FAQ
How much does Vikasit 32B cost?
Vikasit 32B is an open-weight model built on Qwen3-32B (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 32B open weight?
Yes. Vikasit 32B is built on Qwen3-32B (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 32B?
Because Vikasit 32B is open weight, you self-host it with any OpenAI-compatible inference server (such as vLLM or SGLang) loaded with the Qwen3-32B (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 32B support?
Vikasit 32B supports a 32K native, 131K via YaRN context window. It is a 32.8B Dense transformer model — full specifications are listed in the table above.
License & attribution
Apache 2.0
Built on Qwen3-32B (Apache 2.0). Upstream copyright, license, and attribution notices are retained.