Make benchmark comparison grug

Rewrite the shared 27B/35B benchmark section in the model cards' established Grug voice without changing any benchmark values.

README.md

@@ -107,12 +107,12 @@ saving and stability. MATH-500 show the trade win on hard unseen ground.
 - known wart fixed in v2.1: think sometimes too thin on hard problems;
   over-verification in long agent sessions.
 
-## 27b vs 35b on shared benchmarks
+## 27b and 35b hunt same prey
 
-HumanEval and sanitized MBPP are the benchmarks published for both current
-parent checkpoints. Scores are pass@1 percent; bold marks the better result.
+both grug hunt HumanEval and sanitized MBPP. number show pass@1 percent.
+bold grug win that hunt.
 
-| benchmark | [grug-27b](https://huggingface.co/ProCreations/grug-27b) v2.1 | [grug-35b](https://huggingface.co/ProCreations/grug-35b) rebuilt main |
+| hunt | [grug-27b](https://huggingface.co/ProCreations/grug-27b) v2.1 | [grug-35b](https://huggingface.co/ProCreations/grug-35b) rebuilt |
 |---|---:|---:|
 | HumanEval (164) | **87.2** | 80.5 |
 | MBPP sanitized (100) | 85.0 | **88.0** |