Add placeholder MiniMax and GLM vLLM benchmarks (#6068)

* feat: add MiniMax-M2.7 and GLM-5.1 model benchmarks

Validated on p6-b300.48xlarge (8x B300, 2.1TB VRAM):
- minimax-m2.7: TP=4, FP8, 1900 output tok/s → min 950
- glm-5.1: TP=8, bf16, 674 output tok/s → min 330

Fleet 'x86-p6b300-runner' needs to be created in CDK infra.

* chore: comment out MiniMax/GLM entries pending fleet creation

* chore: rename placeholder fleet to x86-p5e-runner

Author: sallyseok

.github/config/model-tests/vllm-model-tests.yml

@@ -104,6 +104,29 @@ benchmark:
       min_throughput: 680
       min_rps: 5.3
 
+    # Pending p5e.48xlarge fleet creation. Fleet name "x86-p5e-runner" is a placeholder.
+    # - name: "minimax-m2.7"
+    #   s3_model: "minimax-m2.7.tar.gz"
+    #   fleet: "x86-p5e-runner"
+    #   extra_args: "--tensor-parallel-size 4 --max-model-len 4096 --dtype auto --trust-remote-code"
+    #   input_len: 512
+    #   output_len: 128
+    #   num_prompts: 64
+    #   batch_size: 4
+    #   min_throughput: 950
+    #   min_rps: 7.4
+
+    # - name: "glm-5.1"
+    #   s3_model: "glm-5.1.tar.gz"
+    #   fleet: "x86-p5e-runner"
+    #   extra_args: "--tensor-parallel-size 8 --max-model-len 4096 --dtype auto --trust-remote-code"
+    #   input_len: 512
+    #   output_len: 128
+    #   num_prompts: 64
+    #   batch_size: 4
+    #   min_throughput: 330
+    #   min_rps: 2.6
+
     - name: "qwen3.5-9b"
       s3_model: "qwen3.5-9b.tar.gz"
       fleet: "x86-g6xl-runner"