README.md

---
license: apache-2.0
base_model:
- Qwen/Qwen3-14B
pipeline_tag: text-generation
---
## NTX-2.1-Pro

NTX-2.1-Pro is a compact 14B-parameter model trained on GPT-5.5 reasoning traces, designed to improve instruction tuning, logical reasoning, and code generation performance while keeping a small footprint.

Highlights:
- 14B parameters, efficient to run vs. larger models
- Distilled from GPT-5.5 traces for reasoning & instruction following
- Strong coding performance for its size class

## Benchmarks

| Benchmark | NTX-2.1-Pro | Qwen3-14B (Thinking) |
|-----------|-------------|------------------------|
| GPQA      | 67.4        | 64.0                  |