diff --git a/config.example.yaml b/config.example.yaml
index b9f0a81..87764cd 100644
--- a/config.example.yaml
+++ b/config.example.yaml
@@ -1,4 +1,3 @@
-concurrency: 1
 shutdown_timeout: 30s
 
 server:
diff --git a/internal/config/config.go b/internal/config/config.go
index 044d5ef..826c50a 100644
--- a/internal/config/config.go
+++ b/internal/config/config.go
@@ -20,9 +20,6 @@ type Config struct {
 	// Address is the host:port to listen on.
 	Address string `yaml:"address"`
 
-	// Concurrency is the number of concurrent requests to allow.
-	// Defaults to 1.
-	Concurrency int `yaml:"concurrency"`
 	// ShutdownTimeout is the maximum time to wait for graceful shutdown.
 	// Defaults to "30s".
 	ShutdownTimeout string `yaml:"shutdown_timeout"`
@@ -42,9 +39,6 @@ type Config struct {
 // ApplyDefaults sets default values for optional configuration fields.
 // Called automatically by Load before validation.
 func (cfg *Config) ApplyDefaults() {
-	if cfg.Concurrency == 0 {
-		cfg.Concurrency = 1
-	}
 	if cfg.ShutdownTimeout == "" {
 		cfg.ShutdownTimeout = "30s"
 	}
diff --git a/internal/config/config_test.go b/internal/config/config_test.go
index 21d2a88..f0367fe 100644
--- a/internal/config/config_test.go
+++ b/internal/config/config_test.go
@@ -11,7 +11,6 @@ import (
 func validConfig() string {
 	return `
 address: ":9090"
-concurrency: 2
 shutdown_timeout: "60s"
 
 llm:
@@ -55,12 +54,6 @@ func TestLoad(t *testing.T) {
 						cfg.Address, ":9090",
 					)
 				}
-				if cfg.Concurrency != 2 {
-					t.Errorf(
-						"Concurrency = %d, want 2",
-						cfg.Concurrency,
-					)
-				}
 				if cfg.ShutdownTimeout != "60s" {
 					t.Errorf(
 						"ShutdownTimeout = %q, want %q",
@@ -96,12 +89,6 @@ tts:
 						cfg.Address, ":8080",
 					)
 				}
-				if cfg.Concurrency != 1 {
-					t.Errorf(
-						"Concurrency = %d, want 1",
-						cfg.Concurrency,
-					)
-				}
 				if cfg.ShutdownTimeout != "30s" {
 					t.Errorf(
 						"ShutdownTimeout = %q, want %q",
@@ -365,9 +352,6 @@ func TestApplyDefaults(t *testing.T) {
 	if cfg.Address != ":8080" {
 		t.Errorf("Address = %q, want %q", cfg.Address, ":8080")
 	}
-	if cfg.Concurrency != 1 {
-		t.Errorf("Concurrency = %d, want 1", cfg.Concurrency)
-	}
 	if cfg.ShutdownTimeout != "30s" {
 		t.Errorf(
 			"ShutdownTimeout = %q, want %q",
@@ -379,7 +363,6 @@ func TestApplyDefaults(t *testing.T) {
 func TestApplyDefaults_NoOverwrite(t *testing.T) {
 	cfg := &Config{
 		Address:         ":9999",
-		Concurrency:     5,
 		ShutdownTimeout: "120s",
 	}
 	cfg.ApplyDefaults()
@@ -387,9 +370,6 @@ func TestApplyDefaults_NoOverwrite(t *testing.T) {
 	if cfg.Address != ":9999" {
 		t.Errorf("Address = %q, want %q", cfg.Address, ":9999")
 	}
-	if cfg.Concurrency != 5 {
-		t.Errorf("Concurrency = %d, want 5", cfg.Concurrency)
-	}
 	if cfg.ShutdownTimeout != "120s" {
 		t.Errorf(
 			"ShutdownTimeout = %q, want %q",
diff --git a/internal/service/service.go b/internal/service/service.go
index eb68192..6acf622 100644
--- a/internal/service/service.go
+++ b/internal/service/service.go
@@ -29,7 +29,6 @@ import (
 
 	"github.com/google/uuid"
 	openai "github.com/sashabaranov/go-openai"
-	"golang.org/x/sync/semaphore"
 )
 
 //go:embed all:static/*
@@ -42,7 +41,6 @@ type Service struct {
 	llm    *llm.Client
 	log    *slog.Logger
 	mux    *http.ServeMux
-	sem    *semaphore.Weighted
 	server *http.Server
 	stt    *stt.Client
 	tmpl   *template.Template
@@ -57,7 +55,6 @@ func New(cfg *config.Config, log *slog.Logger) (*Service, error) {
 		cfg: cfg,
 		log: log,
 		mux: http.NewServeMux(),
-		sem: semaphore.NewWeighted(int64(cfg.Concurrency)),
 	}
 
 	// Setup tool registry.
@@ -193,7 +190,6 @@ func (svc *Service) ServeHTTP(w http.ResponseWriter, r *http.Request) {
 func (svc *Service) Run(ctx context.Context) error {
 	svc.log.Info(
 		"starting odidere",
-		slog.Int("concurrency", svc.cfg.Concurrency),
 		slog.Group(
 			"llm",
 			slog.String("url", svc.cfg.LLM.URL),
@@ -582,17 +578,6 @@ func (svc *Service) voiceStream(w http.ResponseWriter, r *http.Request) {
 		return
 	}
 
-	// Acquire semaphore.
-	if err := svc.sem.Acquire(ctx, 1); err != nil {
-		http.Error(
-			w,
-			"service unavailable",
-			http.StatusServiceUnavailable,
-		)
-		return
-	}
-	defer svc.sem.Release(1)
-
 	var (
 		messages      = req.Messages
 		transcription string