From 61a8b42c001442b947f7b114ed232dfd65764e2e Mon Sep 17 00:00:00 2001
From: Xinyuan Tong <115166877+JustinTong0323@users.noreply.github.com>
Date: Thu, 18 Jun 2026 19:40:56 +0100
Subject: [PATCH] docs(minimax-m3): add MMMU-Pro accuracy to B200 benchmark
 card (#28668)

---
 .../snippets/configs/MiniMaxAI/minimax-m3-benchmarks.jsx  | 2 +-
 docs_new/src/snippets/configs/MiniMaxAI/minimax-m3.jsx    | 8 ++++++++
 2 files changed, 9 insertions(+), 1 deletion(-)

diff --git a/docs_new/src/snippets/configs/MiniMaxAI/minimax-m3-benchmarks.jsx b/docs_new/src/snippets/configs/MiniMaxAI/minimax-m3-benchmarks.jsx
index 32166df7b..beeca9ef1 100644
--- a/docs_new/src/snippets/configs/MiniMaxAI/minimax-m3-benchmarks.jsx
+++ b/docs_new/src/snippets/configs/MiniMaxAI/minimax-m3-benchmarks.jsx
@@ -41,7 +41,7 @@ export const benchmarks = [
       { workload: { dataset: "random", isl: 2048, osl: 256, max_concurrency: 64, num_prompts: 128 },
         ttft_ms: 1580, tpot_ms: 24.1, tokens_per_sec_per_gpu: 265 },
     ],
-    accuracy: { gpqa_pct: 89.1, gsm8k_pct: 96.5 }, // 2026-06-15, sgl-eval --thinking, recommended sampling (temp 1.0/top_p 0.95), tp8. GSM8K full 1319 = 96.51% (greedy 96.89%). GPQA Diamond 198, n-repeats 4 = pass@1[avg-of-4] 89.14% +/-1.73% (pass@4 95.45%, majority@4 93.52%).
+    accuracy: { gpqa_pct: 89.1, gsm8k_pct: 96.5, mmmu_pro_pct: 72.7 }, // 2026-06-15, sgl-eval --thinking, recommended sampling (temp 1.0/top_p 0.95), tp8. GSM8K full 1319 = 96.51% (greedy 96.89%). GPQA Diamond 198, n-repeats 4 = pass@1[avg-of-4] 89.14% +/-1.73% (pass@4 95.45%, majority@4 93.52%). MMMU-Pro 2026-06-18, sgl-eval "standard (10 options)" test split, full 1730, single-shot 72.66% (thinking, temp 1.0/top_p 0.95).
   },
   {
     // Hopper H200: bf16 build (MXFP8 is Blackwell-only) at tp8, built-in Triton
diff --git a/docs_new/src/snippets/configs/MiniMaxAI/minimax-m3.jsx b/docs_new/src/snippets/configs/MiniMaxAI/minimax-m3.jsx
index f54774953..0125fb5a7 100644
--- a/docs_new/src/snippets/configs/MiniMaxAI/minimax-m3.jsx
+++ b/docs_new/src/snippets/configs/MiniMaxAI/minimax-m3.jsx
@@ -71,6 +71,13 @@ sgl-eval run gpqa \\
   --model {{MODEL_NAME}} \\
   --temperature 1.0 --top-p 0.95 \\
   --thinking --n-repeats 4 --max-tokens 40960`,
+      mmmu_pro_pct:
+`pip install git+https://github.com/sgl-project/sgl-eval
+sgl-eval run mmmu_pro \\
+  --base-url http://{{CURL_HOST}}:{{CURL_PORT}}/v1 \\
+  --model {{MODEL_NAME}} \\
+  --temperature 1.0 --top-p 0.95 \\
+  --thinking`,
     },
     numPromptsByConc: { 24: 24, 64: 128 },
   },
@@ -78,6 +85,7 @@ sgl-eval run gpqa \\
   accuracyLabels: [
     ["gpqa_pct", "GPQA Diamond", "%"],
     ["gsm8k_pct", "GSM8K", "%"],
+    ["mmmu_pro_pct", "MMMU-Pro", "%"],
   ],
 
   dockerImages: {