All models
Text / ReasoningAvailable via API1T total · 32B active

Vikasit Titan 1T

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

Overview

Vikasit Titan 1T is a trillion-parameter, natively multimodal agentic MoE — 1T total / 32B active, 384 experts, MLA attention, and a built-in vision encoder. Built for autonomous, long-running agent workflows. Served live via the Vikasit API.

Specifications

Total parameters
1T total
Active parameters
32B active
Architecture
Mixture-of-Experts (MLA), with MoonViT vision encoder
Experts
384 total / 8 selected per token
Layers
61 (incl. 1 dense)
Attention
Multi-head Latent Attention (MLA), SwiGLU
Context window
256K (262K)
Vocabulary
160,000
Modalities
Text + image + video in → text out (natively multimodal)
License
Modified MIT

Capabilities

  • Natively multimodal (text + image + video)
  • Frontier agentic and tool-use performance
  • 256K context, agent-swarm orchestration
  • MLA attention for efficient long context
Multilingual.

Benchmarks

BenchmarkScore
GPQA-Diamond90.5
LiveCodeBench v689.6
SWE-bench Verified80.2
Humanity's Last Exam34.7
Terminal-Bench 2.066.7
AIME 202696.4
MMLU-ProN/A

Numbers from the Kimi K2.6 HuggingFace model card (Moonshot AI). Modified MIT: large-scale commercial deployments above the upstream MAU/revenue threshold must display the original model attribution in the UI.

Hardware & deployment

PrecisionMemory
bf16~2 TB
INT4~500 GB

Quick start

Call Vikasit Titan 1T through the OpenAI-compatible Vikasit AI API at https://api.vikasit.ai/v1 using the model id vikasit-titan-1t.

OpenAI-compatible Python (Vikasit AI API)
# pip install openai
import os
from openai import OpenAI

client = OpenAI(
    base_url="https://api.vikasit.ai/v1",
    api_key=os.environ["VIKASIT_API_KEY"],
)

resp = client.chat.completions.create(
    model="vikasit-titan-1t",
    messages=[
        {"role": "user", "content": "Explain Vikasit Titan 1T in one sentence."}
    ],
)

print(resp.choices[0].message.content)
# or with curl
curl https://api.vikasit.ai/v1/chat/completions \
  -H "Authorization: Bearer $VIKASIT_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "vikasit-titan-1t",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

Limitations

  • Modified-MIT attribution duty at very large scale
  • Cluster-scale serving required

Vikasit Titan 1T FAQ

How much does Vikasit Titan 1T cost?

Vikasit Titan 1T is served through the Vikasit AI API on usage-based, pay-as-you-go pricing billed per million input and output tokens — see the Vikasit AI pricing page for current rates. Because it is built on the open-weight Kimi K2.6 (Moonshot AI, Modified MIT), you can also self-host the weights for free under the Modified MIT licence and pay only for your own compute.

Is Vikasit Titan 1T open weight?

Yes. Vikasit Titan 1T is built on Kimi K2.6 (Moonshot AI, Modified MIT) and distributed under the Modified MIT licence, so the weights are openly available for self-hosting, fine-tuning, and commercial use, subject to the upstream licence terms.

How do I use Vikasit Titan 1T with the OpenAI SDK?

The Vikasit AI API is OpenAI-compatible. Point the OpenAI client's base URL at https://api.vikasit.ai/v1, set your Vikasit API key, and pass "vikasit-titan-1t" as the model. The quick-start snippet above shows the exact Python call.

What context window does Vikasit Titan 1T support?

Vikasit Titan 1T supports a 256K (262K) context window. It is a 1T total (32B active) Mixture-of-Experts (MLA), with MoonViT vision encoder model — full specifications are listed in the table above.

License & attribution

Modified MIT

Built on Kimi K2.6 (Moonshot AI, Modified MIT). Upstream copyright, license, and attribution notices are retained.

Kimi K2.6 model card