diff --git a/include/ircd/gpt/vocab.h b/include/ircd/gpt/vocab.h
index c63d8637e..69147dc9d 100644
--- a/include/ircd/gpt/vocab.h
+++ b/include/ircd/gpt/vocab.h
@@ -42,9 +42,15 @@ namespace ircd::gpt::vocab
 	// Tokenize one token. Error thrown if input is not exactly one token.
 	u16 tokenize(const string_view &in, const bool prefix_space = false);
 
+	// Return the number of tokens which would be output by a tokenize().
+	size_t count(const string_view &in) noexcept;
+
 	// Decode token values to build output text string.
 	string_view detokenize(const mutable_buffer &out, const vector_view<const u16> &in) noexcept;
 
+	// Return the length of the string which would be output by a detokenize().
+	size_t count(const vector_view<const u16> &in) noexcept;
+
 	// Other tools
 	string_view debug(const mutable_buffer &buf, const u16 token, const uint fmt_msk = -1U);
 }
diff --git a/ircd/gpt_vocab.cc b/ircd/gpt_vocab.cc
index a53c2b4c5..bd1ba1540 100644
--- a/ircd/gpt_vocab.cc
+++ b/ircd/gpt_vocab.cc
@@ -213,9 +213,29 @@ ircd::gpt::vocab::debug(const mutable_buffer &out,
 	};
 }
 
-//
-// detokenize
-//
+size_t
+ircd::gpt::vocab::count(const vector_view<const u16> &in)
+noexcept
+{
+	static const size_t max
+	{
+		32_KiB
+	};
+
+	thread_local char buf[max];
+	const auto res
+	{
+		detokenize(buf, in)
+	};
+
+	const auto ret
+	{
+		res.size()
+	};
+
+	assert(ret < max);
+	return ret;
+}
 
 ircd::string_view
 ircd::gpt::vocab::detokenize(const mutable_buffer &out,
@@ -252,9 +272,29 @@ noexcept
 	};
 }
 
-//
-// tokenize
-//
+size_t
+ircd::gpt::vocab::count(const string_view &in)
+noexcept
+{
+	static const size_t max
+	{
+		2048
+	};
+
+	thread_local u16 buf[max];
+	const auto res
+	{
+		tokenize(buf, in)
+	};
+
+	const auto ret
+	{
+		res.size()
+	};
+
+	assert(ret < max);
+	return ret;
+}
 
 uint16_t
 ircd::gpt::vocab::tokenize(const string_view &in,