Multi-domain eval: 5 domains (FLORES, code, math, chat, PDFs) — v2-256k tops the SOTA leaderboard

Full per-language comparison table (all 39 langs, English first)

Expand model card: v1 vs v2 per-language deltas + detail on user-defined tokens for code

OpenEuroLLM Tokenizer v2 (256k) — initial upload

initial commit