# THX-01 Just Processed 70 Million Tokens. Our Open-Source Decision Model Is Just Getting Started.

> Source: <https://dev.to/montydrief/thx-01-just-processed-70-million-tokens-our-open-source-decision-model-is-just-getting-started-2o8l>
> Published: 2026-10-12 09:36:54+00:00

**70 million tokens processed. 322 million parameters. One forward pass. Free API. Fully open-source.**

A few days ago, we open-sourced **THX-01**, our 322M-parameter non-autoregressive decision model designed to make AI decisions faster, more efficient, and accessible to everyone.

Today, we've reached another milestone: **over 70 million tokens processed through our THX-01 API since launch.**

And we're just getting started.

Most AI applications don't actually need to generate long responses.

Sometimes, all you need is a reliable decision:

Instead of generating tokens one by one, THX-01 processes a document or message alongside typed questions and returns structured answers with calibrated probabilities in a single forward pass.

On our GPU benchmark, inference takes approximately **10 milliseconds**.

We evaluated THX-01 against TypeSafe Jev 1.13 on our internal benchmark of 2,843 multilingual support tickets.

| Metric | THX-01 | Jev 1.13 | 
|---|---|---|
| Accuracy | **98.4%** | 97.4% | 
| Calibration Error (ECE) | **0.003** | 0.007 | 
| Measured Latency | **~10 ms** | 331 ms | 
| Open-Source Weights | Yes | No | 

THX-01 outperformed Jev on this classification benchmark, with approximately 33x lower measured latency.

The latency figures come from different serving environments, so they should not be interpreted as a hardware-controlled comparison.

Our results show that small, specialized decision models can deliver competitive performance without requiring billions of additional parameters.

THX-01 is trained using **Reinforcement Learning for Calibrated Decisions (RLCD)**.

Rather than optimizing only for correct predictions, RLCD encourages probability estimates that reflect the model's actual uncertainty.

This is particularly useful for AI agents and production applications that need to decide whether to accept a prediction, request additional information, or escalate to another model.

THX-01 supports:

The model supports 18 post-training languages, including Azerbaijani, English, Russian, and Turkish.

And at just 322M parameters, it can even run locally on smartphones with a compatible inference runtime.

We believe developers should be able to experiment with new AI architectures without worrying about infrastructure costs.

That's why we've opened the **THX-01 API with a free tier of 200 decisions per minute**.

You can test it with a simple request:

```
curl https://api.hal-x.ai/v1/systemone \
  -H "Content-Type: application/json" \
  -d '{
    "state": "My card was charged twice for one order",
    "questions": {
      "team": {
        "type": "choice",
        "question": "Which team should handle this?",
        "criteria": {
          "billing": "Billing issues",
          "technical": "Technical issues",
          "sales": "Sales inquiries"
        }
      }
    }
  }'
```

Check the API documentation for the current authentication requirements and usage limits.

**API Documentation:** [https://api.hal-x.ai/docs/thx-01](https://api.hal-x.ai/docs/thx-01)

We released THX-01 under the **Apache 2.0 license**.

Developers can access the model weights, source code, and training resources on Hugging Face.

You can also install the Python package directly:

```
pip install thx01
```

**Model & Source Code:** [https://huggingface.co/doofz/THX-01](https://huggingface.co/doofz/THX-01)

**Interactive Demo:** [https://huggingface.co/spaces/doofz/THX-01-demo](https://huggingface.co/spaces/doofz/THX-01-demo)

At HAL-X AI, we're working toward a future where AI systems don't just generate answers, but make reliable, measurable, and well-calibrated decisions.

We believe the next generation of AI infrastructure will need specialized models that are smaller, faster, and capable of expressing uncertainty.

**70 million tokens processed is an exciting milestone for our team, but this is only the beginning.**

We're continuing to improve THX-01, expand its capabilities, and contribute to the global open-source AI ecosystem.

Our mission is simple:

**We need more AI models that know how to make decisions, how confident they should be, and when they don't know.**

And we're building them for everyone.

**Try THX-01 today:**

`pip install thx01`