README: tested environments, Blackwell note

README.md

@@ -109,7 +109,12 @@ curl -s localhost:30171/v1/systemone -H 'content-type: application/json' -d '{
 ```
 
 The server batches concurrent requests (`--max-batch-requests`, `--max-batch-tokens`) and binds `127.0.0.1` by
-default. It runs on one GPU with the backbone in bfloat16 and the head in float32. For CPU or Apple-silicon use, see
+default. It runs on one GPU with the backbone in bfloat16 and the head in float32.
+
+Tested with this exact quick start on an NVIDIA H200 (PyTorch 2.11, CUDA 12.8) and on AMD MI350X (PyTorch 2.9, ROCm 7.2);
+the JevBench public tiers reproduce the numbers above. On Blackwell GPUs (B200/B300) install a CUDA 13 build of PyTorch
+first (`pip install torch --index-url https://download.pytorch.org/whl/cu130`); the CUDA 12.8 build lacks kernels for
+those GPUs. For CPU or Apple-silicon use, see
 [StandardOne-8B-SH-GGUF](https://huggingface.co/StandardThinking/StandardOne-8B-SH-GGUF).
 
 ## Files

SHA256SUMS

@@ -1,5 +1,5 @@
 1495e1e757ef4d0925a2350563cf5754bb23c51701a8ec4fb3c5cdcbedae6747  LICENSE
-2e46a8977e057ca9c452138915ceda5dd9fa946143d0b5c9ce0feb6f59925efc  README.md
+d9ac54c2ca89376a28d11ddd398a69ca10690371dcb2013f2402c9d7cd497e62  README.md
 66e0fda8c8ab269ccd010dca6196931d475a22c3f2eebe5b6b3a2ec33a354d0e  SYSTEM_PROMPT.txt
 e62f05da25555a4c109d0f413eb673d328fcc40f7c82537c61631611f6954372  chat_template.jinja
 5f3d6035ca36885c9be04b7cfa88b5ba532c87d5b26fa3525c94327fb1b4348b  code/head.py