Update README.md

README.md

@@ -19,9 +19,9 @@ tags:
 
 # BCLM-1-Small
 
-We are releasing **BCLM-1**, an open family of flagship language models that deliver strong, predictable performance from a compact parameter budget.
+We are releasing **BCLM-1**, an open family of flagship language models that deliver strong, predictable performance across a range of tasks.
 
-BCLM-1 is built on a hybrid SSM/attention architecture (Bulldog) for efficient training and inference. It handles writing, summarization, and recall tasks out of the box, holding onto information across its full context window where similarly sized models fall apart.
+BCLM-1 is built on a hybrid SSM/attention architecture (Bulldog) for efficient training and inference, excelling at writing, summarization, and recall out of the box.
 
 ## Models
 
@@ -68,7 +68,12 @@ Transformers support coming soon.
 
 ## Training
 
-BCLM-1-Small was pretrained for 24,000 steps, reaching a training loss of 2.10 (perplexity 8). The subsequent SFT stage ran for 6,000 steps, achieving a validation loss of 1.84.
+Trained on NVIDIA H200 GPUs.
+
+| Stage | Steps | Loss | Perplexity (ppl) |
+|---|---|---|---|
+| Pretraining | 24,000 | 2.11 | 8.25 |
+| SFT | 6,000 | 1.85 | 6.36 |
 
 ## Tokenizer and Supported Languages