| --- |
| 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 | |
