diff --git a/src/check_builtin.cpp b/src/check_builtin.cpp
index 9a3013e8e..92ee2299f 100644
--- a/src/check_builtin.cpp
+++ b/src/check_builtin.cpp
@@ -4594,6 +4594,13 @@ gb_internal bool check_builtin_procedure(CheckerContext *c, Operand *operand, As
 				operand->value.value_float = bit_cast<f64>(abs);
 				break;
 			}
+			case ExactValue_Rational: {
+				mp_int n; mp_init(&n);
+				defer (mp_clear(&n));
+				mp_abs(&operand->value.value_rational->num, &n);
+				operand->value = exact_value_rational_from_ints(&n, &operand->value.value_rational->den);
+				break;
+			}
 			case ExactValue_Complex: {
 				f64 r = operand->value.value_complex->real;
 				f64 i = operand->value.value_complex->imag;
diff --git a/src/check_expr.cpp b/src/check_expr.cpp
index 8c75dd8c0..9de4d8859 100644
--- a/src/check_expr.cpp
+++ b/src/check_expr.cpp
@@ -4873,6 +4873,11 @@ gb_internal void check_binary_expr(CheckerContext *c, Operand *x, Ast *node, Typ
 					fail = true;
 				}
 				break;
+			case ExactValue_Rational:
+				if (big_int_is_zero(&y->value.value_rational->num)) {
+					fail = true;
+				}
+				break;
 			}
 
 			if (fail) {
diff --git a/src/checker.cpp b/src/checker.cpp
index fa65098a8..59f1c9c25 100644
--- a/src/checker.cpp
+++ b/src/checker.cpp
@@ -1480,6 +1480,7 @@ gb_internal void init_universal(void) {
 			type = t_untyped_integer;
 			break;
 		case ExactValue_Float:
+		case ExactValue_Rational:
 			type = t_untyped_float;
 			break;
 		}
diff --git a/src/exact_value.cpp b/src/exact_value.cpp
index 60126a401..1cd282c2a 100644
--- a/src/exact_value.cpp
+++ b/src/exact_value.cpp
@@ -502,7 +502,7 @@ gb_internal ExactValue exact_value_to_integer(ExactValue v) {
 		return exact_value_i64(cast(i64)cast(intptr)v.value_pointer);
 
 	case ExactValue_Rational:
-		// Only an exact integer (den == 1 after reduction) converts to an integer.
+		// NOTE(bill): Only an exact integer (den == 1 after reduction) converts to an integer
 		if (mp_cmp_d(&v.value_rational->den, 1) == MP_EQ) {
 			ExactValue r = {ExactValue_Integer};
 			r.value_integer = {0};
@@ -1339,6 +1339,14 @@ gb_internal gbString write_exact_value_to_string(gbString str, ExactValue const
 	case ExactValue_Float:
 		return gb_string_append_fmt(str, "%.17g", v.value_float);
 	case ExactValue_Rational:
+		// Integer-valued (den == 1, e.g. an overflowing literal like `1.0e400`) prints its exact decimal,
+		// so a diagnostic shows the real magnitude rather than an f64 that has rounded to +Inf.
+		if (mp_cmp_d(&v.value_rational->den, 1) == MP_EQ) {
+			String s = big_int_to_string(heap_allocator(), &v.value_rational->num);
+			str = gb_string_append_length(str, s.text, s.len);
+			gb_free(heap_allocator(), s.text);
+			return str;
+		}
 		return gb_string_append_fmt(str, "%.17g", big_rat_to_f64(&v.value_rational->num, &v.value_rational->den));
 	case ExactValue_Complex:
 		return gb_string_append_fmt(str, "%.17g+%.17gi", v.value_complex->real, v.value_complex->imag);