All models
Text / ReasoningOpen weights27B

Vikasit 27B

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

Overview

Vikasit 27B is a powerful dense model from the 3.6 generation — deep reasoning, advanced coding, and research-grade analysis with a 262K native context window.

Specifications

Total parameters
27B
Architecture
Dense (hybrid Gated Attention + Gated DeltaNet)
Layers
64
Attention
Gated Attention (24 query / 4 KV heads) + Gated DeltaNet layers
Context window
262K native, ~1M via YaRN
Modalities
Text in → text out (multimodal-capable base)
License
Apache 2.0

Capabilities

  • Deep reasoning and research analysis
  • Advanced coding with strong SWE-bench
  • 262K native context, ~1M via YaRN
  • Hybrid linear+full attention efficiency
Multilingual. Strong English + major Indian languages.

Benchmarks

BenchmarkScore
MMLU-Pro86.2
GPQA-Diamond87.8
LiveCodeBench v683.9
SWE-bench Verified77.2
AIME 202694.1
MATH-500N/A
IFEvalN/A

Numbers from the Qwen3.6-27B HuggingFace model card. AIME 2026 reported instead of 2025.

Hardware & deployment

PrecisionMemory
bf16~54 GB
INT4~16 GB

Quick start

Vikasit 27B is an open-weight model. Self-host it with any OpenAI-compatible inference server and call it with the OpenAI SDK as shown below.

OpenAI-compatible Python (self-hosted, e.g. vLLM)
# 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-27b",
    messages=[
        {"role": "user", "content": "Explain Vikasit 27B in one sentence."}
    ],
)

print(resp.choices[0].message.content)

Limitations

  • Dense compute cost above sibling MoE models
  • IFEval/MATH-500 not on official card

Vikasit 27B FAQ

How much does Vikasit 27B cost?

Vikasit 27B is an open-weight model built on Qwen3.6-27B (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 27B open weight?

Yes. Vikasit 27B is built on Qwen3.6-27B (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 27B?

Because Vikasit 27B is open weight, you self-host it with any OpenAI-compatible inference server (such as vLLM or SGLang) loaded with the Qwen3.6-27B (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 27B support?

Vikasit 27B supports a 262K native, ~1M via YaRN context window. It is a 27B Dense (hybrid Gated Attention + Gated DeltaNet) model — full specifications are listed in the table above.

License & attribution

Apache 2.0

Built on Qwen3.6-27B (Apache 2.0). Upstream copyright, license, and attribution notices are retained.

Qwen3.6-27B model card