TORUS is a SnapKitty finance-focused language model fine-tuned for financial research, structured analysis, and production workflows. TORUS is designed to operate with retrieval-augmented financial knowledge and WORM-chain audit infrastructure, providing a model layer for traceable research, financial intelligence, and production-grade agent workflows.

Built for integration with SnapKitty Agents, TORUS combines specialized financial inference with externally managed retrieval, immutable recordkeeping, and auditable execution pipelines. The model is intended for environments where financial outputs need to be connected to source material, preserved through tamper-resistant records, and integrated into repeatable production processes.

Model Card

Overview

TORUS is a text language model distributed through Ollama.

The published Ollama artifact is:

jessicalw34/TORUS:latest

TORUS uses the same published specifications as the SnapKitty model family:

• 2.7 GB
• 4K context window
• Text modality
• Ollama runtime

Run TORUS

ollama run jessicalw34/TORUS

API

curl http://localhost:11434/api/chat \
  -d '{
    "model": "jessicalw34/TORUS",
    "messages": [
      {
        "role": "user",
        "content": "Hello!"
      }
    ]
  }'

Python

from ollama import chat

response = chat(
    model="jessicalw34/TORUS",
    messages=[
        {
            "role": "user",
            "content": "Hello!"
        }
    ],
)

print(response.message.content)

JavaScript

import ollama from "ollama";

const response = await ollama.chat({
  model: "jessicalw34/TORUS",
  messages: [
    {
      role: "user",
      content: "Hello!"
    }
  ]
});

console.log(response.message.content);

Model Architecture

The public model card does not independently specify:

• Base model
• Parameter count
• Training dataset
• Fine-tuning method
• Quantization format
• Training hardware
• Training duration
• Evaluation benchmarks

Those details should be documented when formally released.

Context

4K context window

Applications should design prompts and agent context within the available context limit.

Modality

Text

Agent Integration

TORUS can serve as the inference component of a SnapKitty agent runtime.

┌─────────────────────────┐
│   SnapKitty Agent       │
├─────────────────────────┤
│ Objective               │
│ State                   │
│ Tools                   │
│ Permissions             │
│ Execution               │
└────────────┬────────────┘
             │
             ▼
┌─────────────────────────┐
│          TORUS          │
│      Language Model     │
└────────────┬────────────┘
             │
             ▼
┌─────────────────────────┐
│         Ollama          │
└─────────────────────────┘

The model and agent runtime are separate components.

TORUS provides model inference. The surrounding SnapKitty agent system determines how model output is interpreted and whether external tools or actions are permitted.

Limitations

TORUS may produce incorrect or incomplete outputs.

Model output should be validated before being used for consequential decisions or external actions.

Connecting TORUS to an agent runtime does not itself establish correctness, autonomy, or tool authorization.

Reproducibility

Model:
jessicalw34/TORUS

Tag:
latest

Runtime:
Ollama

Size:
2.7 GB

Context:
4K

Modality:
Text

Pin an immutable model digest when reproducibility is required.

Versioning

Recommended release structure:

TORUS:1
TORUS:1.1
TORUS:2

Each release should document changes to:

• Model weights
• Runtime configuration
• Context length
• Quantization
• Prompt configuration
• Training or fine-tuning
• Evaluation

Evaluation

Benchmark results should be reported against the exact TORUS release tested.

Future evaluations should identify:

• Exact TORUS version
• Benchmark
• Prompt configuration
• Context length
• Sampling parameters
• Hardware
• Runtime version

Sovereign Source

This model is a SnapKitty sovereign-source artifact.

Copyright © 2026 SnapKitty.

The model artifact, Modelfile, configuration, prompts, and associated source materials are governed by the license accompanying this release.

No rights are granted beyond those expressly provided by that license. Redistribution, modification, commercial use, or creation of derivative artifacts is subject to the applicable license terms.

For the authoritative terms visit:

collectivekitty.com

Token Policy

Preview Tier

All newly issued API keys begin on the Preview tier.

• 10,000 research tokens per day
• Daily allowance resets at 00:00 UTC
• 50,000 research tokens lifetime maximum
• Lifetime cap is hard enforced
• Upgrade to the Standard tier is required after the lifetime allocation is exhausted

Limit Enforcement

When a token limit is reached:

HTTP 429 Too Many Requests

Daily:

{
  "error": "daily_token_limit_reached",
  "message": "Daily token limit reached (10,000 tokens). Resets at midnight UTC."
}

Lifetime:

{
  "error": "lifetime_token_limit_reached",
  "message": "Preview token limit reached (50,000 tokens). Upgrade to standard tier."
}

Citation

When referencing this model:

jessicalw34/TORUS

Identify the exact version or digest where reproducibility matters.

TORUS

2.7 GB · 4K context · Text

Ollama model:

jessicalw34/TORUS:latest

Run:

ollama run jessicalw34/TORUS