diff --git a/internal/metal/batch.go b/internal/metal/batch.go
index 3c5c04e..103003f 100644
--- a/internal/metal/batch.go
+++ b/internal/metal/batch.go
@@ -3,11 +3,11 @@
 package metal
 
 import (
+	"cmp"
 	"context"
 	"fmt"
 	"math"
 	"slices"
-	"sort"
 	"time"
 )
 
@@ -49,8 +49,8 @@ func (m *Model) Classify(ctx context.Context, prompts []string, cfg GenerateConf
 	for i := range indices {
 		indices[i] = i
 	}
-	sort.Slice(indices, func(a, b int) bool {
-		return lengths[indices[a]] > lengths[indices[b]]
+	slices.SortFunc(indices, func(a, b int) int {
+		return cmp.Compare(lengths[b], lengths[a])
 	})
 
 	maxLen := lengths[indices[0]]
@@ -160,8 +160,8 @@ func (m *Model) BatchGenerate(ctx context.Context, prompts []string, cfg Generat
 	for i := range indices {
 		indices[i] = i
 	}
-	sort.Slice(indices, func(a, b int) bool {
-		return lengths[indices[a]] > lengths[indices[b]]
+	slices.SortFunc(indices, func(a, b int) int {
+		return cmp.Compare(lengths[b], lengths[a])
 	})
 
 	N := int32(len(prompts))
@@ -208,7 +208,7 @@ func (m *Model) BatchGenerate(ctx context.Context, prompts []string, cfg Generat
 		maxTokens = 256
 	}
 
-	for step := 0; step < maxTokens; step++ {
+	for step := range maxTokens {
 		select {
 		case <-ctx.Done():
 			// Return partial results on cancellation.
diff --git a/internal/metal/generate.go b/internal/metal/generate.go
index 342d095..066cc64 100644
--- a/internal/metal/generate.go
+++ b/internal/metal/generate.go
@@ -4,6 +4,7 @@ package metal
 
 import (
 	"context"
+	"errors"
 	"fmt"
 	"iter"
 	"slices"
@@ -185,7 +186,7 @@ func (m *Model) Generate(ctx context.Context, prompt string, cfg GenerateConfig)
 			Free(logits)
 		}()
 
-		for i := 0; i < cfg.MaxTokens; i++ {
+		for i := range cfg.MaxTokens {
 			select {
 			case <-ctx.Done():
 				m.lastErr = ctx.Err()
@@ -256,7 +257,7 @@ func (m *Model) Generate(ctx context.Context, prompt string, cfg GenerateConfig)
 func (m *Model) InspectAttention(ctx context.Context, prompt string) (*AttentionResult, error) {
 	tokens := m.tokenizer.Encode(prompt)
 	if len(tokens) == 0 {
-		return nil, fmt.Errorf("empty prompt after tokenisation")
+		return nil, errors.New("empty prompt after tokenisation")
 	}
 
 	caches := m.newCaches()
@@ -304,7 +305,7 @@ func (m *Model) InspectAttention(ctx context.Context, prompt string) (*Attention
 
 		keys[i] = make([][]float32, numHeads)
 		stride := validLen * headDim
-		for h := 0; h < numHeads; h++ {
+		for h := range numHeads {
 			start := h * stride
 			end := start + stride
 			if end > len(flat) {
diff --git a/internal/metal/grad.go b/internal/metal/grad.go
index 485a432..b08ea9f 100644
--- a/internal/metal/grad.go
+++ b/internal/metal/grad.go
@@ -39,7 +39,7 @@ func goGradFunc(outputs *C.mlx_vector_array, inputs C.mlx_vector_array, payload
 
 	nInputs := int(C.mlx_vector_array_size(inputs))
 	goInputs := make([]*Array, nInputs)
-	for i := 0; i < nInputs; i++ {
+	for i := range nInputs {
 		a := newArray("GRAD_INPUT")
 		C.mlx_vector_array_get(&a.ctx, inputs, C.size_t(i))
 		goInputs[i] = a
@@ -324,7 +324,7 @@ func MSELoss(predictions, targets *Array) *Array {
 func vectorToArrays(vec C.mlx_vector_array) []*Array {
 	n := int(C.mlx_vector_array_size(vec))
 	out := make([]*Array, n)
-	for i := 0; i < n; i++ {
+	for i := range n {
 		a := newArray("VEC_OUT")
 		C.mlx_vector_array_get(&a.ctx, vec, C.size_t(i))
 		out[i] = a
diff --git a/internal/metal/lora.go b/internal/metal/lora.go
index 8d080b2..2bcb661 100644
--- a/internal/metal/lora.go
+++ b/internal/metal/lora.go
@@ -12,10 +12,11 @@ import (
 	"encoding/json"
 	"fmt"
 	"log/slog"
+	"maps"
 	"math"
 	"os"
 	"path/filepath"
-	"sort"
+	"slices"
 	"strings"
 	"unsafe"
 )
@@ -152,12 +153,7 @@ func (a *LoRAAdapter) TotalParams() int {
 
 // SortedNames returns layer names in deterministic sorted order.
 func (a *LoRAAdapter) SortedNames() []string {
-	names := make([]string, 0, len(a.Layers))
-	for name := range a.Layers {
-		names = append(names, name)
-	}
-	sort.Strings(names)
-	return names
+	return slices.Sorted(maps.Keys(a.Layers))
 }
 
 // AllTrainableParams returns all trainable arrays (A and B from every layer),
diff --git a/internal/metal/slice.go b/internal/metal/slice.go
index 4e8b4af..8a6928d 100644
--- a/internal/metal/slice.go
+++ b/internal/metal/slice.go
@@ -31,7 +31,7 @@ func SliceAxis(a *Array, axis int, start, end int32) *Array {
 	ndim := a.NumDims()
 	starts := make([]int32, ndim)
 	ends := make([]int32, ndim)
-	for i := 0; i < ndim; i++ {
+	for i := range ndim {
 		starts[i] = 0
 		ends[i] = int32(a.Dim(i))
 	}
diff --git a/internal/metal/tokenizer.go b/internal/metal/tokenizer.go
index 6f2c1f4..ea798dc 100644
--- a/internal/metal/tokenizer.go
+++ b/internal/metal/tokenizer.go
@@ -191,7 +191,7 @@ func (t *Tokenizer) bpeMerge(symbols []string) []string {
 		// Find the pair with the lowest merge rank.
 		bestRank := -1
 		bestIdx := -1
-		for i := 0; i < len(symbols)-1; i++ {
+		for i := range len(symbols) - 1 {
 			key := symbols[i] + " " + symbols[i+1]
 			if rank, ok := t.mergeRanks[key]; ok {
 				if bestRank < 0 || rank < bestRank {