diff --git a/c/doctor.py b/c/doctor.py
index da64704..4cb5a42 100644
--- a/c/doctor.py
+++ b/c/doctor.py
@@ -48,7 +48,7 @@ def run_doctor(model, ram_gb=0, context=4096, gpu_indices=None, vram_gb=0, *,
 
     config = model / "config.json"
     try:
-        valid_config = isinstance(json.loads(config.read_text()), dict)
+        valid_config = isinstance(json.loads(config.read_text(encoding="utf-8")), dict)
     except (OSError, ValueError):
         valid_config = False
     checks.append(_check("model.config", "pass" if valid_config else "fail",
diff --git a/c/resource_plan.py b/c/resource_plan.py
index 33674e1..2fe4c3c 100644
--- a/c/resource_plan.py
+++ b/c/resource_plan.py
@@ -39,7 +39,7 @@ def analyze_model(model):
     config_path = model / "config.json"
     if not config_path.is_file():
         raise ValueError(f"missing config.json: {model}")
-    config = json.loads(config_path.read_text())
+    config = json.loads(config_path.read_text(encoding="utf-8"))
     shards = sorted(model.glob("*.safetensors"))
     if not shards:
         raise ValueError(f"no safetensors shards: {model}")
@@ -210,15 +210,15 @@ def build_plan(model, ram_gb=0, context=4096, gpu_indices=None, vram_gb=0,
     if ram_budget < 4 * GB:
         ram_budget = 8 * GB
     typical = info["typical_expert_bytes"]
-    layers = int(cfg.get("num_hidden_layers", 0)) + 1
-    kv_bytes = layers * context * (int(cfg.get("kv_lora_rank", 0)) +
-                                   int(cfg.get("qk_rope_head_dim", 0))) * 4
-    kv_buffer = context * int(cfg.get("num_attention_heads", 0)) * (
-        int(cfg.get("qk_nope_head_dim", 0)) + int(cfg.get("v_head_dim", 0))) * 4
+    layers = int(cfg.get("num_hidden_layers") or 0) + 1
+    kv_bytes = layers * context * (int(cfg.get("kv_lora_rank") or 0) +
+                                   int(cfg.get("qk_rope_head_dim") or 0)) * 4
+    kv_buffer = context * int(cfg.get("num_attention_heads") or 0) * (
+        int(cfg.get("qk_nope_head_dim") or 0) + int(cfg.get("v_head_dim") or 0)) * 4
     runtime_bytes = int(1.2 * GB + 2.5 * GB + 64 * typical + kv_bytes + kv_buffer)
     cache_bytes = max(0, ram_budget - info["dense_bytes"] - runtime_bytes)
     per_cap = info["per_cap_bytes"]
-    configured_experts = int(cfg.get("n_routed_experts", 0))
+    configured_experts = int(cfg.get("n_routed_experts") or 0)
     cap = int(cache_bytes // per_cap) if per_cap else 0
     if configured_experts:
         cap = min(cap, configured_experts)