[skip ci] Update nightly support matrices

Signed-off-by: Buildkite Bot <buildkite-bot@users.noreply.github.com>

Author: buildkite-bot

support_matrices/nightly/feature_support_matrix.csv

@@ -1,11 +1,11 @@
 Feature,CorrectnessTest,PerformanceTest
-"Collective Communication Matmul",✅,N/A
 "Prefix Caching",✅,✅
-"Multimodal Inputs",✅,✅
-"Quantized Matmul Attention and KV Cache",✅,✅
 "Chunked Prefill",✅,✅
-"JAX-Path Qxix Quantization",✅,✅
-"Single Program Multi Data",✅,✅
-"Speculative Decoding: Ngram",✅,✅
 "Structured Decoding",✅,N/A
+"Single Program Multi Data",✅,✅
+"Quantized Matmul Attention and KV Cache",✅,✅
+"JAX-Path Qxix Quantization",✅,✅
+"Multimodal Inputs",✅,✅
+"Speculative Decoding: Ngram",✅,❌
+"Collective Communication Matmul",✅,N/A
 "Ragged Paged Attention V3",✅,✅

support_matrices/nightly/model_support_matrix.csv

@@ -1,8 +1,8 @@
 Model,UnitTest,IntegrationTest,Benchmark
 "meta-llama/Llama-3.3-70B-Instruct",✅,✅,✅
+"Qwen/Qwen3-4B",✅,✅,✅
 "Qwen/Qwen2.5-VL-7B-Instruct",✅,❌,N/A
-"Qwen/Qwen3-32B",✅,✅,✅
 "google/gemma-3-27b-it",✅,✅,✅
-"meta-llama/Llama-3.1-8B-Instruct",✅,✅,✅
+"Qwen/Qwen3-32B",✅,✅,✅
+"meta-llama/Llama-3.1-8B-Instruct",✅,✅,❌
 "Qwen/Qwen3-30B-A3B",✅,❌,N/A
-"Qwen/Qwen3-4B",✅,✅,✅