Vikasit Reasoner 1T
Trillion-scale reasoning MoE. Deep multi-step reasoning, long-horizon agents.
Overview
Vikasit Reasoner 1T is a trillion-parameter reasoning MoE tuned for deep, multi-step reasoning and long-horizon agentic tasks. Configurable reasoning effort (high / extra-high). Served live via the Vikasit API.
Specifications
- Total parameters
- ~1T total
- Active parameters
- ~63B active
- Architecture
- Mixture-of-Experts (reasoning-tuned)
- Context window
- 128K native, 256K via YaRN
- Modalities
- Text in → text out
- License
- MIT
Capabilities
- Deep chain-of-thought reasoning
- Long-horizon agentic planning
- Configurable reasoning effort (high / xhigh)
- Strong abstract / ARC-style reasoning
Benchmarks
| Benchmark | Score |
|---|---|
| GPQA-Diamond | 88.3 |
| SWE-bench Verified | 74.0 |
| AIME 2026 | 95.8 |
| ARC-AGI-V2 | 66.2 |
| MMLU-Pro | N/A |
| LiveCodeBench | N/A |
Numbers from the Ring-2.6-1T HuggingFace model card (inclusionAI / Ant Group). Architecture internals (expert/layer counts) are not published on the official card; active-param figure is vendor/secondary.
Hardware & deployment
| Precision | Memory |
|---|---|
| bf16 | ~2 TB |
| INT4 | ~500 GB |
Quick start
Call Vikasit Reasoner 1T through the OpenAI-compatible Vikasit AI API at https://api.vikasit.ai/v1 using the model id vikasit-reasoner-1t.
# 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-reasoner-1t",
messages=[
{"role": "user", "content": "Explain Vikasit Reasoner 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-reasoner-1t",
"messages": [{"role": "user", "content": "Hello"}]
}'Limitations
- Text-only; no multimodal
- Architecture internals not disclosed by base
- Vendor-reported benchmarks; limited third-party reproduction
Vikasit Reasoner 1T FAQ
How much does Vikasit Reasoner 1T cost?
Vikasit Reasoner 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 Ring-2.6-1T (inclusionAI / Ant Group, MIT), you can also self-host the weights for free under the MIT licence and pay only for your own compute.
Is Vikasit Reasoner 1T open weight?
Yes. Vikasit Reasoner 1T is built on Ring-2.6-1T (inclusionAI / Ant Group, MIT) and distributed under the 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 Reasoner 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-reasoner-1t" as the model. The quick-start snippet above shows the exact Python call.
What context window does Vikasit Reasoner 1T support?
Vikasit Reasoner 1T supports a 128K native, 256K via YaRN context window. It is a ~1T total (~63B active) Mixture-of-Experts (reasoning-tuned) model — full specifications are listed in the table above.
License & attribution
MIT
Built on Ring-2.6-1T (inclusionAI / Ant Group, MIT). Upstream copyright, license, and attribution notices are retained.