diff --git a/include/cxxopts.hpp b/include/cxxopts.hpp
index 891e78a..dfc0976 100644
--- a/include/cxxopts.hpp
+++ b/include/cxxopts.hpp
@@ -27,6 +27,7 @@ THE SOFTWARE.
 #ifndef CXXOPTS_HPP_INCLUDED
 #define CXXOPTS_HPP_INCLUDED
 
+#include <cerrno>
 #include <cstdint>
 #include <cstdlib>
 #include <cstring>
@@ -999,6 +1000,27 @@ check_signed_range(bool negative, U value, const std::string& text)
   SignedCheck<T, std::numeric_limits<T>::is_signed>()(negative, value, text);
 }
 
+template <typename T>
+void
+floating_point_parser(const std::string& text, T& value, T (*parser)(const char*, char**))
+{
+  if (text.empty())
+  {
+    throw_or_mimic<exceptions::incorrect_argument_type>(text);
+  }
+
+  char* end = nullptr;
+  errno = 0;
+  const auto result = parser(text.c_str(), &end);
+
+  if (end == text.c_str() || errno == ERANGE)
+  {
+    throw_or_mimic<exceptions::incorrect_argument_type>(text);
+  }
+
+  value = result;
+}
+
 } // namespace detail
 
 template <typename R, typename T>
@@ -1132,6 +1154,27 @@ parse_value(const std::string& text, std::string& value)
   value = text;
 }
 
+inline
+void
+parse_value(const std::string& text, float& value)
+{
+  detail::floating_point_parser(text, value, std::strtof);
+}
+
+inline
+void
+parse_value(const std::string& text, double& value)
+{
+  detail::floating_point_parser(text, value, std::strtod);
+}
+
+inline
+void
+parse_value(const std::string& text, long double& value)
+{
+  detail::floating_point_parser(text, value, std::strtold);
+}
+
 // The fallback parser. It uses the stringstream parser to parse all types
 // that have not been overloaded explicitly.  It has to be placed in the
 // source code before all other more specialized templates.
diff --git a/test/options.cpp b/test/options.cpp
index fb495ed..cda65c1 100644
--- a/test/options.cpp
+++ b/test/options.cpp
@@ -1,4 +1,5 @@
 #include "catch.hpp"
+#include <cmath>
 #include <iostream>
 
 #include <initializer_list>
@@ -977,6 +978,52 @@ TEST_CASE("Floats", "[options]")
   CHECK(positional[3] == -1.5e6);
 }
 
+TEST_CASE("Special floating point values", "[options]")
+{
+  cxxopts::Options options("parses_special_floats", "parses special floating point values");
+  options.add_options()
+    ("double", "Double precision", cxxopts::value<double>())
+    ("long-double", "Extended precision", cxxopts::value<long double>())
+    ("positional", "Floats", cxxopts::value<std::vector<float>>());
+
+  Argv av({"special_floats", "--double", "inf", "--long-double", "infinity", "--", "-inf", "nan"});
+
+  auto** argv = av.argv();
+  auto argc = av.argc();
+
+  options.parse_positional("positional");
+  auto result = options.parse(argc, argv);
+
+  auto parsed_double = result["double"].as<double>();
+  CHECK(std::isinf(parsed_double));
+  CHECK(!std::signbit(parsed_double));
+
+  auto parsed_long_double = result["long-double"].as<long double>();
+  CHECK(std::isinf(parsed_long_double));
+  CHECK(!std::signbit(parsed_long_double));
+
+  auto& positional = result["positional"].as<std::vector<float>>();
+  REQUIRE(positional.size() == 2);
+  CHECK(std::isinf(positional[0]));
+  CHECK(std::signbit(positional[0]));
+  CHECK(std::isnan(positional[1]));
+}
+
+TEST_CASE("Invalid floats", "[options]")
+{
+  cxxopts::Options options("invalid_floats", "rejects invalid floats");
+  options.add_options()
+    ("positional", "Floats", cxxopts::value<std::vector<float>>());
+
+  Argv av({"floats", "--", "abc"});
+
+  auto** argv = av.argv();
+  auto argc = av.argc();
+
+  options.parse_positional("positional");
+  CHECK_THROWS_AS(options.parse(argc, argv), cxxopts::exceptions::incorrect_argument_type);
+}
+
 TEST_CASE("Invalid integers", "[integer]") {
     cxxopts::Options options("invalid_integers", "rejects invalid integers");
     options.add_options()