From 38f7c287957d8d6203d2bbaa1c370fd6634b4d31 Mon Sep 17 00:00:00 2001
From: Sascha Rogmann <github@rogmann.org>
Date: Fri, 2 Jan 2026 00:10:46 +0100
Subject: [PATCH] server: can_speculate() tests self-spec

---
 tools/server/server-context.cpp | 5 +++--
 1 file changed, 3 insertions(+), 2 deletions(-)

diff --git a/tools/server/server-context.cpp b/tools/server/server-context.cpp
index 5d492479db..7a72fa98e4 100644
--- a/tools/server/server-context.cpp
+++ b/tools/server/server-context.cpp
@@ -258,8 +258,9 @@ struct server_slot {
         return state != SLOT_STATE_IDLE;
     }
 
+    // Checks if a draft model is active or self-speculation using context-tokens
     bool can_speculate() const {
-        return ctx_dft;
+        return ctx_dft || task->params.speculative.use_self;
     }
 
     void add_token(const completion_token_output & token) {
@@ -274,7 +275,7 @@ struct server_slot {
     int get_n_draft_max() const {
         GGML_ASSERT(task);
 
-        if (!can_speculate() && !task->params.speculative.use_self) {
+        if (!can_speculate()) {
             return 0;
         }