Enable benchmarked two-slot medium profile

config/install.env.example

@@ -82,7 +82,7 @@ FAST_UBATCH_SIZE=32
 MEDIUM_CONTEXT=160000
 MEDIUM_BATCH_SIZE=2048
 MEDIUM_UBATCH_SIZE=128
-MEDIUM_TENSOR_SPLIT=90,10
+MEDIUM_TENSOR_SPLIT=85,15
 LARGE_CONTEXT=192000
 LARGE_BATCH_SIZE=2048
 LARGE_UBATCH_SIZE=128
@@ -99,6 +99,12 @@ UNCENSORED_MTP_MAX=2
 EXPERIMENTAL_CONTEXT=76800
 LLAMA_THREADS=6
 LLAMA_THREADS_BATCH=6
+FAST_PARALLEL_SLOTS=1
+MEDIUM_PARALLEL_SLOTS=2
+LARGE_PARALLEL_SLOTS=1
+ULTRA_PARALLEL_SLOTS=1
+UNCENSORED_PARALLEL_SLOTS=1
+EXPERIMENTAL_PARALLEL_SLOTS=1
 PIPER_TTS_VERSION=1.6.0
 PIPER_VOICE=de_DE-thorsten-high
 XTTS_IMAGE=ghcr.io/coqui-ai/xtts-streaming-server:latest-cuda121@sha256:f7fb3b1f9d4bc88af94da1b5959d8002f1e0b003c97557164034eb8a29f01b90

config/profile-matrix.json

@@ -7,6 +7,7 @@
       "id": "fast",
       "alias": "qwen-fast",
       "context": 76800,
+      "parallel_slots": 1,
       "model_env": "FAST_MODEL_FILE",
       "model_family": "Qwen3.8-27B IQ4 Mix",
       "gpu_split": "5080 only",
@@ -18,9 +19,10 @@
       "id": "medium",
       "alias": "qwen-medium",
       "context": 160000,
+      "parallel_slots": 2,
       "model_env": "MEDIUM_MODEL_FILE",
       "model_family": "Qwen3.8-27B IQ4 XS Pure",
-      "gpu_split": "90:10",
+      "gpu_split": "85:15",
       "vision": true,
       "mtp": 3,
       "description": "Ausgewogenes Standardprofil für Alltag und lange agentische Aufgaben."
@@ -29,6 +31,7 @@
       "id": "large",
       "alias": "qwen-large",
       "context": 192000,
+      "parallel_slots": 1,
       "model_env": "LARGE_MODEL_FILE",
       "model_family": "Qwen3.8-27B IQ4 XS Pure",
       "gpu_split": "86:14",
@@ -40,6 +43,7 @@
       "id": "ultra",
       "alias": "qwen-ultra",
       "context": 262144,
+      "parallel_slots": 1,
       "model_env": "ULTRA_MODEL_FILE",
       "model_family": "Qwen3.8-27B IQ4 XS Pure",
       "gpu_split": "80:20",
@@ -51,6 +55,7 @@
       "id": "uncensored",
       "alias": "qwen-uncensored",
       "context": 80000,
+      "parallel_slots": 1,
       "model_env": "UNCENSORED_MODEL_FILE",
       "model_family": "Qwen3.8-27B Abliterated Q4_K_M",
       "gpu_split": "90:10",