diff --git a/src/uu/od/src/prn_float.rs b/src/uu/od/src/prn_float.rs
index 155ce7d07..c93b02d25 100644
--- a/src/uu/od/src/prn_float.rs
+++ b/src/uu/od/src/prn_float.rs
@@ -37,8 +37,61 @@ pub static FORMAT_ITEM_BF16: FormatterItemInfo = FormatterItemInfo {
     formatter: FormatWriter::BFloatWriter(format_item_bf16),
 };
 
+/// Clean up a normalized float string by removing unnecessary padding and digits.
+/// - Strip leading spaces.
+/// - Trim trailing zeros after the decimal point (and the dot itself if empty).
+/// - Leave the exponent part (e/E...) untouched.
+fn trim_float_repr(raw: &str) -> String {
+    // Drop padding added by `format!` width specification
+    let mut s = raw.trim_start().to_string();
+
+    // Keep NaN/Inf representations as-is
+    let lower = s.to_ascii_lowercase();
+    if lower == "nan" || lower == "inf" || lower == "-inf" {
+        return s;
+    }
+
+    // Separate exponent from mantissa
+    let mut exp_part = String::new();
+    if let Some(idx) = s.find(['e', 'E']) {
+        exp_part = s[idx..].to_string();
+        s.truncate(idx);
+    }
+
+    // Trim trailing zeros in mantissa, then remove trailing dot if left alone
+    if s.contains('.') {
+        while s.ends_with('0') {
+            s.pop();
+        }
+        if s.ends_with('.') {
+            s.pop();
+        }
+    }
+
+    // If everything was trimmed, leave a single zero
+    if s.is_empty() || s == "-" || s == "+" {
+        s.push('0');
+    }
+
+    s.push_str(&exp_part);
+    s
+}
+
+/// Pad a floating value to a fixed width for column alignment while keeping
+/// the original precision (including trailing zeros). This mirrors the
+/// behavior of other float formatters (`f32`, `f64`) and keeps the output
+/// stable across platforms.
+fn pad_float_repr(raw: &str, width: usize) -> String {
+    format!("{raw:>width$}")
+}
+
 pub fn format_item_f16(f: f64) -> String {
-    format!(" {}", format_f16(f16::from_f64(f)))
+    let value = f16::from_f64(f);
+    let width = FORMAT_ITEM_F16.print_width - 1;
+    // Format once, trim redundant zeros, then re-pad to the canonical width
+    let raw = format_f16(value);
+    let trimmed = trim_float_repr(&raw);
+    format!(" {}", pad_float_repr(&trimmed, width))
 }
 
 pub fn format_item_f32(f: f64) -> String {
@@ -82,7 +135,10 @@ fn format_f64_exp_precision(f: f64, width: usize, precision: usize) -> String {
 
 pub fn format_item_bf16(f: f64) -> String {
     let bf = bf16::from_f32(f as f32);
-    format!(" {}", format_binary16_like(f, 15, 8, is_subnormal_bf16(bf)))
+    let width = FORMAT_ITEM_BF16.print_width - 1;
+    let raw = format_binary16_like(f64::from(bf), width, 8, is_subnormal_bf16(bf));
+    let trimmed = trim_float_repr(&raw);
+    format!(" {}", pad_float_repr(&trimmed, width))
 }
 
 fn format_f16(f: f16) -> String {
diff --git a/tests/by-util/test_od.rs b/tests/by-util/test_od.rs
index 54be34551..fea019e3a 100644
--- a/tests/by-util/test_od.rs
+++ b/tests/by-util/test_od.rs
@@ -3,7 +3,7 @@
 // For the full copyright and license information, please view the LICENSE
 // file that was distributed with this source code.
 
-// spell-checker:ignore abcdefghijklmnopqrstuvwxyz Anone fdbb littl
+// spell-checker:ignore abcdefghijklmnopqrstuvwxyz Anone fdbb littl bfloat
 
 #[cfg(unix)]
 use std::io::Read;
@@ -197,6 +197,32 @@ fn test_hex32() {
         .stdout_only(expected_output);
 }
 
+// Regression: 16-bit IEEE half should print with canonical precision (no spurious digits)
+#[test]
+fn test_float16_compact() {
+    let input: [u8; 4] = [0x3c, 0x00, 0x3c, 0x00]; // two times 1.0 in big-endian half
+    new_ucmd!()
+        .arg("--endian=big")
+        .arg("-An")
+        .arg("-tfH")
+        .run_piped_stdin(&input[..])
+        .success()
+        .stdout_only("               1               1\n");
+}
+
+// Regression: 16-bit bfloat should print with canonical precision (no spurious digits)
+#[test]
+fn test_bfloat16_compact() {
+    let input: [u8; 4] = [0x3f, 0x80, 0x3f, 0x80]; // two times 1.0 in big-endian bfloat16
+    new_ucmd!()
+        .arg("--endian=big")
+        .arg("-An")
+        .arg("-tfB")
+        .run_piped_stdin(&input[..])
+        .success()
+        .stdout_only("               1               1\n");
+}
+
 #[test]
 fn test_f16() {
     let input: [u8; 14] = [
@@ -210,7 +236,7 @@ fn test_f16() {
     ]; // 0x8400 -6.104e-5
     let expected_output = unindent(
         "
-            0000000       1.0000000               0              -0             inf
+            0000000               1               0              -0             inf
             0000010            -inf             NaN   -6.1035156e-5
             0000016
             ",
@@ -237,7 +263,7 @@ fn test_fh() {
     ]; // 0x8400 -6.1035156e-5
     let expected_output = unindent(
         "
-            0000000       1.0000000               0              -0             inf
+            0000000               1               0              -0             inf
             0000010            -inf             NaN   -6.1035156e-5
             0000016
         ",
@@ -264,7 +290,7 @@ fn test_fb() {
     ]; // -6.1035156e-5
     let expected_output = unindent(
         "
-            0000000       1.0000000               0              -0             inf
+            0000000               1               0              -0             inf
             0000010            -inf             NaN   -6.1035156e-5
             0000016
         ",