PROMETECH Cicikuş Classic upgrades GPT-2 reasoning

This is a small-model flex, not a frontier-model moment. The real question is whether the fine-tuning stack actually extracts useful reasoning from a legacy GPT-2 base or just wraps it in ambitious branding.

• –Built on `openai-community/gpt2-medium`, so the novelty is in the data and training recipe, not the backbone.
• –The 0.4B footprint makes it interesting for local inference, low-cost experiments, and embedded use cases where bigger models are overkill.
• –The BCE framing reads as product narrative as much as technical differentiator; outside benchmarks would matter a lot here.
• –Bilingual Turkish-English support is the most concrete advantage if it holds up on real instruction tasks.
• –The MIT model license is attractive, but the separate BCE tech/licensing language suggests commercial users should read carefully.