Sync model card from docs/model_cards/qwen14b-code-trainer-v8_mixed.md

Phase 4 benchmark: gsm8k (0-shot)

V7 agent eval (nexus-aligned): 5/5 making progress

V7 tool-calling eval (nexus-aligned): 6/10 (60%)

Phase 4 eval-only: phase4-eval-v7.json

Phase 4A v8_mixed sweep — eval_loss=0.4837215840816498

Phase 4 benchmark: gsm8k (0-shot)

Phase 4 eval-only: phase4-eval-v7.json

V7 agent eval (nexus-aligned): 5/5 making progress

V7 tool-calling eval (nexus-aligned): 2/10 (20%)

Phase 4A v8_mixed sweep — eval_loss=0.4584397077560425

initial commit