diff --git a/crates/xtrain-train/src/bin/train.rs b/crates/xtrain-train/src/bin/train.rs
index dcbe059..22ff49f 100644
--- a/crates/xtrain-train/src/bin/train.rs
+++ b/crates/xtrain-train/src/bin/train.rs
@@ -167,6 +167,20 @@ fn main() {
         }
     });
 
+    // Eval-only mode: load a checkpoint and score it on the held-out val set, then
+    // exit. Used to put an EXISTING model (e.g. v0) and a new one on the same
+    // metric — the v0-vs-v1 val-loss comparison. The arch flags must match the ckpt.
+    if let Some(p) = args.iter().position(|a| a == "--eval-ckpt") {
+        let ckpt_path = PathBuf::from(args.get(p + 1).expect("--eval-ckpt <path>"));
+        xtrain_train::checkpoint::load_into(&ckpt_path, &model.params())
+            .expect("load eval checkpoint");
+        let v = valid.expect("--eval-ckpt needs --val-tokens > 0");
+        let vl = xtrain_train::eval_loss(&model, device, &v, seq_len, eval_batches);
+        println!("eval-only: {} → val loss {vl:.4}", ckpt_path.display());
+        sample_some(&model, device, &tok_path);
+        return;
+    }
+
     let tcfg = TrainConfig {
         seq_len,
         batch_size,
diff --git a/crates/xtrain-train/src/lib.rs b/crates/xtrain-train/src/lib.rs
index 6a9c1d1..2c00d1d 100644
--- a/crates/xtrain-train/src/lib.rs
+++ b/crates/xtrain-train/src/lib.rs
@@ -19,4 +19,4 @@ pub mod sample;
 mod train_loop;
 
 #[cfg(not(no_cuda))]
-pub use train_loop::{TrainConfig, TrainResult, train};
+pub use train_loop::{TrainConfig, TrainResult, eval_loss, train};
diff --git a/crates/xtrain-train/src/train_loop.rs b/crates/xtrain-train/src/train_loop.rs
index 0acbcf3..0185e5a 100644
--- a/crates/xtrain-train/src/train_loop.rs
+++ b/crates/xtrain-train/src/train_loop.rs
@@ -153,8 +153,8 @@ pub fn train(
 
 /// Mean cross-entropy over `batches` deterministic, non-overlapping windows of
 /// the validation corpus (no backward — eval only). Deterministic so val loss is
-/// comparable across steps and runs.
-fn eval_loss(
+/// comparable across steps and runs (and across models — the v0-vs-v1 metric).
+pub fn eval_loss(
     model: &TinyTransformer,
     device: Device,
     valid: &Corpus,