diff --git a/core/bifrost.go b/core/bifrost.go
index e72d9c9723..f745ef6b67 100644
--- a/core/bifrost.go
+++ b/core/bifrost.go
@@ -1135,7 +1135,18 @@ func (bifrost *Bifrost) tryStreamRequest(req *schemas.BifrostRequest, ctx contex
 		bifrost.releaseChannelMessage(msg)
 		return stream, nil
 	case bifrostErrVal := <-msg.Err:
+		bifrost.logger.Warn("error while executing stream request: %v", bifrostErrVal.Error.Message)
+		// Marking final chunk
+		ctx = context.WithValue(ctx, schemas.BifrostContextKeyStreamEndIndicator, true)
+		// On error we will complete post-hooks
+		recoveredResp, recoveredErr := pipeline.RunPostHooks(&ctx, nil, &bifrostErrVal, len(bifrost.plugins))
 		bifrost.releaseChannelMessage(msg)
+		if recoveredErr != nil {
+			return nil, recoveredErr
+		}
+		if recoveredResp != nil {
+			return newBifrostMessageChan(recoveredResp), nil
+		}
 		return nil, &bifrostErrVal
 	}
 }
diff --git a/core/changelog.md b/core/changelog.md
index a0e9b0d602..5ccb752e6a 100644
--- a/core/changelog.md
+++ b/core/changelog.md
@@ -1,2 +1,4 @@
 <!-- The pattern we follow here is to keep the changelog for the latest version -->
-<!-- Old changelogs are automatically attached to the GitHub releases -->
\ No newline at end of file
+<!-- Old changelogs are automatically attached to the GitHub releases -->
+
+- Fix: Updates token calculation for streaming responses. #520
\ No newline at end of file
diff --git a/core/providers/openai.go b/core/providers/openai.go
index 9bae458024..2533456539 100644
--- a/core/providers/openai.go
+++ b/core/providers/openai.go
@@ -499,9 +499,24 @@ func handleOpenAIStreaming(
 			}
 
 			// Handle usage-only chunks (when stream_options include_usage is true)
-			if len(response.Choices) == 0 && response.Usage != nil {
+			if response.Usage != nil {
 				// Collect usage information and send at the end of the stream
-				usage = response.Usage
+				// Here in some cases usage comes before final message
+				// So we need to check if the response.Usage is nil and then if usage != nil
+				// then add up all tokens
+				if response.Usage.PromptTokens > usage.PromptTokens {
+					usage.PromptTokens = response.Usage.PromptTokens
+				}
+				if response.Usage.CompletionTokens > usage.CompletionTokens {
+					usage.CompletionTokens = response.Usage.CompletionTokens
+				}
+				if response.Usage.TotalTokens > usage.TotalTokens {
+					usage.TotalTokens = response.Usage.TotalTokens
+				}
+				calculatedTotal := usage.PromptTokens + usage.CompletionTokens
+				if calculatedTotal > usage.TotalTokens {
+					usage.TotalTokens = calculatedTotal
+				}
 				response.Usage = nil
 			}
 
diff --git a/core/version b/core/version
index 36638c8584..54eae6b4d8 100644
--- a/core/version
+++ b/core/version
@@ -1 +1 @@
-1.1.37
+1.1.38
diff --git a/framework/changelog.md b/framework/changelog.md
index 57e331e113..aeca76e348 100644
--- a/framework/changelog.md
+++ b/framework/changelog.md
@@ -1,4 +1,4 @@
 <!-- The pattern we follow here is to keep the changelog for the latest version -->
 <!-- Old changelogs are automatically attached to the GitHub releases -->
 
-- Pricing module now accommodates nested model names i.e. groq/openai/gpt-oss-20b was getting skipped while computing costs.
\ No newline at end of file
+- upgrade: core upgrades to 1.1.38
\ No newline at end of file
diff --git a/framework/logstore/sqlite.go b/framework/logstore/sqlite.go
index e09ec7cf45..39f4f3de37 100644
--- a/framework/logstore/sqlite.go
+++ b/framework/logstore/sqlite.go
@@ -179,7 +179,7 @@ func (s *SQLiteLogStore) SearchLogs(filters SearchFilters, pagination Pagination
 		}
 		return nil, err
 	}
-
+	
 	return &SearchResult{
 		Logs:       logs,
 		Pagination: pagination,
diff --git a/framework/logstore/tables.go b/framework/logstore/tables.go
index a815142ec4..6e6286963c 100644
--- a/framework/logstore/tables.go
+++ b/framework/logstore/tables.go
@@ -233,6 +233,7 @@ func (l *Log) SerializeFields() error {
 	}
 
 	if l.ErrorDetailsParsed != nil {
+		l.ErrorDetailsParsed.Error.Error = nil
 		if data, err := json.Marshal(l.ErrorDetailsParsed); err != nil {
 			return err
 		} else {
diff --git a/framework/version b/framework/version
index 154b9fce5b..79728fe87f 100644
--- a/framework/version
+++ b/framework/version
@@ -1 +1 @@
-1.0.23
+1.0.24
diff --git a/plugins/governance/changelog.md b/plugins/governance/changelog.md
index 6dcfe4eddc..c47f6fa2b6 100644
--- a/plugins/governance/changelog.md
+++ b/plugins/governance/changelog.md
@@ -1,4 +1,5 @@
 <!-- The pattern we follow here is to keep the changelog for the latest version -->
 <!-- Old changelogs are automatically attached to the GitHub releases -->
 
-- Upgrades framework to 1.0.23
\ No newline at end of file
+- upgrade: core to 1.1.38
+- upgrade: framework to 1.0.24
\ No newline at end of file
diff --git a/plugins/governance/version b/plugins/governance/version
index f69752ab13..b66183a60b 100644
--- a/plugins/governance/version
+++ b/plugins/governance/version
@@ -1 +1 @@
-1.2.16
+1.2.17
diff --git a/plugins/jsonparser/changelog.md b/plugins/jsonparser/changelog.md
index 6dcfe4eddc..c47f6fa2b6 100644
--- a/plugins/jsonparser/changelog.md
+++ b/plugins/jsonparser/changelog.md
@@ -1,4 +1,5 @@
 <!-- The pattern we follow here is to keep the changelog for the latest version -->
 <!-- Old changelogs are automatically attached to the GitHub releases -->
 
-- Upgrades framework to 1.0.23
\ No newline at end of file
+- upgrade: core to 1.1.38
+- upgrade: framework to 1.0.24
\ No newline at end of file
diff --git a/plugins/jsonparser/version b/plugins/jsonparser/version
index 05060b805b..a96f385f15 100644
--- a/plugins/jsonparser/version
+++ b/plugins/jsonparser/version
@@ -1 +1 @@
-1.2.15
\ No newline at end of file
+1.2.16
\ No newline at end of file
diff --git a/plugins/logging/changelog.md b/plugins/logging/changelog.md
index 06359c9de0..e12a27e83b 100644
--- a/plugins/logging/changelog.md
+++ b/plugins/logging/changelog.md
@@ -1,5 +1,6 @@
 <!-- The pattern we follow here is to keep the changelog for the latest version -->
 <!-- Old changelogs are automatically attached to the GitHub releases -->
 
-- Upgrades framework to 1.0.23
-- Fixes pricing computation for nested model names.
\ No newline at end of file
+- fix: fixes error logging for streaming and non-streaming responses.
+- upgrade: core to 1.1.38
+- upgrade: framework to 1.0.24
\ No newline at end of file
diff --git a/plugins/logging/main.go b/plugins/logging/main.go
index fbeb3e2b7b..57f6e5330d 100644
--- a/plugins/logging/main.go
+++ b/plugins/logging/main.go
@@ -371,7 +371,7 @@ func (p *LoggerPlugin) PostHook(ctx *context.Context, result *schemas.BifrostRes
 	if !ok {
 		p.logger.Error("model not found in context")
 		return result, err, nil
-	}	
+	}
 	// Check if this is a streaming response
 	requestType, ok := (*ctx).Value(schemas.BifrostContextKeyRequestType).(schemas.RequestType)
 	if !ok {
diff --git a/plugins/logging/operations.go b/plugins/logging/operations.go
index 4a344c1712..7d9173f330 100644
--- a/plugins/logging/operations.go
+++ b/plugins/logging/operations.go
@@ -37,7 +37,7 @@ func (p *LoggerPlugin) updateLogEntry(ctx context.Context, requestID string, tim
 	updates := make(map[string]interface{})
 	if !timestamp.IsZero() {
 		// Try to get original timestamp from context first for latency calculation
-		latency, err := p.calculateLatency(requestID, timestamp, ctx)
+		latency, err := p.calculateLatency(ctx, requestID, timestamp)
 		if err != nil {
 			return err
 		}
@@ -142,7 +142,7 @@ func (p *LoggerPlugin) processStreamUpdate(ctx context.Context, requestID string
 
 	// Handle error case first
 	if data.ErrorDetails != nil {
-		latency, err := p.calculateLatency(requestID, timestamp, ctx)
+		latency, err := p.calculateLatency(ctx, requestID, timestamp)
 		if err != nil {
 			// If we can't get created_at, just update status and error
 			tempEntry := &logstore.Log{}
@@ -182,7 +182,7 @@ func (p *LoggerPlugin) processStreamUpdate(ctx context.Context, requestID string
 	if isFinalChunk {
 		// Stream is finishing, calculate latency
 		var err error
-		latency, err = p.calculateLatency(requestID, timestamp, ctx)
+		latency, err = p.calculateLatency(ctx, requestID, timestamp)
 		if err != nil {
 			return fmt.Errorf("failed to get created_at for latency calculation: %w", err)
 		}
@@ -267,14 +267,17 @@ func (p *LoggerPlugin) processStreamUpdate(ctx context.Context, requestID string
 }
 
 // calculateLatency computes latency in milliseconds from creation time
-func (p *LoggerPlugin) calculateLatency(requestID string, currentTime time.Time, ctx context.Context) (float64, error) {
+func (p *LoggerPlugin) calculateLatency(ctx context.Context, requestID string, currentTime time.Time) (float64, error) {
 	// Try to get original timestamp from context first
 	if ctxTimestamp, ok := ctx.Value(CreatedTimestampKey).(time.Time); ok {
 		return float64(currentTime.Sub(ctxTimestamp).Nanoseconds()) / 1e6, nil
 	}
-
-	// Fallback to database query if not found in context
-	originalEntry, err := p.store.FindFirst(map[string]interface{}{"id": requestID}, "created_at")
+	var originalEntry *logstore.Log
+	err := retryOnNotFound(ctx, func() error {
+		var opErr error
+		originalEntry, opErr = p.store.FindFirst(map[string]interface{}{"id": requestID}, "created_at")
+		return opErr
+	})
 	if err != nil {
 		return 0, err
 	}
diff --git a/plugins/logging/streaming.go b/plugins/logging/streaming.go
index 206f8822ce..4cdaac72a3 100644
--- a/plugins/logging/streaming.go
+++ b/plugins/logging/streaming.go
@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"time"
 
+	"github.com/bytedance/sonic"
 	bifrost "github.com/maximhq/bifrost/core"
 	"github.com/maximhq/bifrost/core/schemas"
 	"github.com/maximhq/bifrost/framework/logstore"
@@ -125,7 +126,7 @@ func (p *LoggerPlugin) addStreamChunk(requestID string, chunk *StreamChunk, obje
 }
 
 // processAccumulatedChunks processes all accumulated chunks in order
-func (p *LoggerPlugin) processAccumulatedChunks(requestID string) error {
+func (p *LoggerPlugin) processAccumulatedChunks(ctx context.Context, requestID string, respErr *schemas.BifrostError) error {
 	accumulator := p.getOrCreateStreamAccumulator(requestID)
 
 	accumulator.mu.Lock()
@@ -138,7 +139,7 @@ func (p *LoggerPlugin) processAccumulatedChunks(requestID string) error {
 	completeMessage := p.buildCompleteMessageFromChunks(accumulator.Chunks)
 
 	// Calculate final latency
-	latency, err := p.calculateLatency(requestID, accumulator.FinalTimestamp, context.Background())
+	latency, err := p.calculateLatency(ctx, requestID, accumulator.FinalTimestamp)
 	if err != nil {
 		p.logger.Error("failed to calculate latency for request %s: %v", requestID, err)
 		latency = 0
@@ -147,6 +148,9 @@ func (p *LoggerPlugin) processAccumulatedChunks(requestID string) error {
 	// Update database with complete message
 	updates := make(map[string]interface{})
 	updates["status"] = "success"
+	if respErr != nil {
+		updates["status"] = "error"
+	}
 	updates["stream"] = true
 	updates["latency"] = latency
 	updates["timestamp"] = accumulator.FinalTimestamp
@@ -158,13 +162,19 @@ func (p *LoggerPlugin) processAccumulatedChunks(requestID string) error {
 	if completeMessage.AssistantMessage != nil && completeMessage.AssistantMessage.ToolCalls != nil {
 		tempEntry.ToolCallsParsed = completeMessage.AssistantMessage.ToolCalls
 	}
-
 	if err := tempEntry.SerializeFields(); err != nil {
 		return fmt.Errorf("failed to serialize complete message: %w", err)
 	}
-
-	updates["output_message"] = tempEntry.OutputMessage
-	updates["content_summary"] = tempEntry.ContentSummary
+	if respErr != nil {
+		if b, mErr := sonic.Marshal(respErr); mErr == nil {
+			updates["error_details"] = string(b)
+		} else {
+			updates["error_details"] = fmt.Sprintf(`{"message":"failed to marshal error: %v"}`, mErr)
+		}
+	} else {
+		updates["output_message"] = tempEntry.OutputMessage
+		updates["content_summary"] = tempEntry.ContentSummary
+	}
 	if tempEntry.ToolCalls != "" {
 		updates["tool_calls"] = tempEntry.ToolCalls
 	}
@@ -402,7 +412,7 @@ func (p *LoggerPlugin) handleStreamingResponse(ctx *context.Context, result *sch
 
 			if shouldProcess {
 
-				if processErr := p.processAccumulatedChunks(requestID); processErr != nil {
+				if processErr := p.processAccumulatedChunks(*ctx, requestID, err); processErr != nil {
 					p.logger.Error("failed to process accumulated chunks for request %s: %v", requestID, processErr)
 				}
 
diff --git a/plugins/maxim/changelog.md b/plugins/maxim/changelog.md
index 6dcfe4eddc..c47f6fa2b6 100644
--- a/plugins/maxim/changelog.md
+++ b/plugins/maxim/changelog.md
@@ -1,4 +1,5 @@
 <!-- The pattern we follow here is to keep the changelog for the latest version -->
 <!-- Old changelogs are automatically attached to the GitHub releases -->
 
-- Upgrades framework to 1.0.23
\ No newline at end of file
+- upgrade: core to 1.1.38
+- upgrade: framework to 1.0.24
\ No newline at end of file
diff --git a/plugins/maxim/version b/plugins/maxim/version
index 95b25aee25..3336003dcc 100644
--- a/plugins/maxim/version
+++ b/plugins/maxim/version
@@ -1 +1 @@
-1.3.6
+1.3.7
diff --git a/plugins/mocker/changelog.md b/plugins/mocker/changelog.md
index 6dcfe4eddc..c47f6fa2b6 100644
--- a/plugins/mocker/changelog.md
+++ b/plugins/mocker/changelog.md
@@ -1,4 +1,5 @@
 <!-- The pattern we follow here is to keep the changelog for the latest version -->
 <!-- Old changelogs are automatically attached to the GitHub releases -->
 
-- Upgrades framework to 1.0.23
\ No newline at end of file
+- upgrade: core to 1.1.38
+- upgrade: framework to 1.0.24
\ No newline at end of file
diff --git a/plugins/mocker/version b/plugins/mocker/version
index fd9d1a5aca..1fc5b8206a 100644
--- a/plugins/mocker/version
+++ b/plugins/mocker/version
@@ -1 +1 @@
-1.2.14
+1.2.15
diff --git a/plugins/semanticcache/changelog.md b/plugins/semanticcache/changelog.md
index 6dcfe4eddc..c47f6fa2b6 100644
--- a/plugins/semanticcache/changelog.md
+++ b/plugins/semanticcache/changelog.md
@@ -1,4 +1,5 @@
 <!-- The pattern we follow here is to keep the changelog for the latest version -->
 <!-- Old changelogs are automatically attached to the GitHub releases -->
 
-- Upgrades framework to 1.0.23
\ No newline at end of file
+- upgrade: core to 1.1.38
+- upgrade: framework to 1.0.24
\ No newline at end of file
diff --git a/plugins/semanticcache/version b/plugins/semanticcache/version
index 591bdbcd6a..9ddbe16da5 100644
--- a/plugins/semanticcache/version
+++ b/plugins/semanticcache/version
@@ -1 +1 @@
-1.2.18
+1.2.19
diff --git a/plugins/telemetry/changelog.md b/plugins/telemetry/changelog.md
index 6dcfe4eddc..c47f6fa2b6 100644
--- a/plugins/telemetry/changelog.md
+++ b/plugins/telemetry/changelog.md
@@ -1,4 +1,5 @@
 <!-- The pattern we follow here is to keep the changelog for the latest version -->
 <!-- Old changelogs are automatically attached to the GitHub releases -->
 
-- Upgrades framework to 1.0.23
\ No newline at end of file
+- upgrade: core to 1.1.38
+- upgrade: framework to 1.0.24
\ No newline at end of file
diff --git a/plugins/telemetry/version b/plugins/telemetry/version
index 05060b805b..a96f385f15 100644
--- a/plugins/telemetry/version
+++ b/plugins/telemetry/version
@@ -1 +1 @@
-1.2.15
\ No newline at end of file
+1.2.16
\ No newline at end of file
diff --git a/tests/core-providers/openrouter_test.go b/tests/core-providers/openrouter_test.go
index 8ce5ca342a..ea371b9cfc 100644
--- a/tests/core-providers/openrouter_test.go
+++ b/tests/core-providers/openrouter_test.go
@@ -1,7 +1,6 @@
 package tests
 
 import (
-	"os"
 	"testing"
 
 	"github.com/maximhq/bifrost/tests/core-providers/config"
@@ -10,9 +9,6 @@ import (
 )
 
 func TestOpenRouter(t *testing.T) {
-	if os.Getenv("OPENROUTER_API_KEY") == "" {
-		t.Skip("OPENROUTER_API_KEY not set; skipping OpenRouter tests")
-	}
 	client, ctx, cancel, err := config.SetupTest()
 	if err != nil {
 		t.Fatalf("Error initializing test setup: %v", err)
@@ -44,3 +40,71 @@ func TestOpenRouter(t *testing.T) {
 
 	runAllComprehensiveTests(t, client, ctx, testConfig)
 }
+
+// TestOpenRouterAnthropic tests Anthropic models via OpenRouter
+func TestOpenRouterAnthropic(t *testing.T) {
+	client, ctx, cancel, err := config.SetupTest()
+	if err != nil {
+		t.Fatalf("Error initializing test setup: %v", err)
+	}
+	defer cancel()
+	defer client.Shutdown()
+
+	testConfig := config.ComprehensiveTestConfig{
+		Provider:  schemas.OpenRouter,
+		ChatModel: "anthropic/claude-3.5-sonnet", // Using Claude 3.5 Sonnet via OpenRouter
+		TextModel: "", // Anthropic models don't support text completion
+		EmbeddingModel: "",
+		Scenarios: config.TestScenarios{
+			TextCompletion:        false, // Not supported by Anthropic
+			SimpleChat:            true,
+			ChatCompletionStream:  true,
+			MultiTurnConversation: true,
+			ToolCalls:             true,
+			MultipleToolCalls:     true,
+			End2EndToolCalling:    true,
+			AutomaticFunctionCall: true,
+			ImageURL:              true,
+			ImageBase64:           true,
+			MultipleImages:        true,
+			CompleteEnd2End:       true,
+			ProviderSpecific:      false, // Skip provider-specific tests
+		},
+	}
+
+	runAllComprehensiveTests(t, client, ctx, testConfig)
+}
+
+// TestOpenRouterMetaLlama tests Meta's Llama models via OpenRouter
+func TestOpenRouterMetaLlama(t *testing.T) {
+	client, ctx, cancel, err := config.SetupTest()
+	if err != nil {
+		t.Fatalf("Error initializing test setup: %v", err)
+	}
+	defer cancel()
+	defer client.Shutdown()
+
+	testConfig := config.ComprehensiveTestConfig{
+		Provider:  schemas.OpenRouter,
+		ChatModel: "meta-llama/llama-3.1-70b-instruct", // Using Llama 3.1 70B via OpenRouter
+		TextModel: "meta-llama/llama-3.1-8b-instruct",  // Using smaller model for text completion
+		EmbeddingModel: "",
+		Scenarios: config.TestScenarios{
+			TextCompletion:        true,
+			SimpleChat:            true,
+			ChatCompletionStream:  true,
+			MultiTurnConversation: true,
+			ToolCalls:             true,
+			MultipleToolCalls:     true,
+			End2EndToolCalling:    true,
+			AutomaticFunctionCall: true,
+			ImageURL:              false, // Llama models typically don't support image inputs
+			ImageBase64:           false,
+			MultipleImages:        false,
+			CompleteEnd2End:       true,
+			ProviderSpecific:      false,
+		},
+	}
+
+	runAllComprehensiveTests(t, client, ctx, testConfig)
+}
\ No newline at end of file
diff --git a/tests/core-providers/scenarios/chat_completion_stream.go b/tests/core-providers/scenarios/chat_completion_stream.go
index 6af2e6c3ba..672a5fd85f 100644
--- a/tests/core-providers/scenarios/chat_completion_stream.go
+++ b/tests/core-providers/scenarios/chat_completion_stream.go
@@ -46,6 +46,7 @@ func RunChatCompletionStreamTest(t *testing.T, client *bifrost.Bifrost, ctx cont
 		var fullContent strings.Builder
 		var responseCount int
 		var lastResponse *schemas.BifrostStream
+		var hasReceivedUsage bool
 
 		// Create a timeout context for the stream reading
 		streamCtx, cancel := context.WithTimeout(ctx, 30*time.Second)
@@ -97,6 +98,22 @@ func RunChatCompletionStreamTest(t *testing.T, client *bifrost.Bifrost, ctx cont
 					}
 				}
 
+				// Check if this response contains usage information
+				if response.Usage != nil {
+					hasReceivedUsage = true
+					t.Logf("📊 Token usage received - Prompt: %d, Completion: %d, Total: %d",
+						response.Usage.PromptTokens,
+						response.Usage.CompletionTokens,
+						response.Usage.TotalTokens)
+
+					// Validate token counts
+					assert.Greater(t, response.Usage.PromptTokens, 0, "Prompt tokens should be greater than 0")
+					assert.Greater(t, response.Usage.CompletionTokens, 0, "Completion tokens should be greater than 0")
+					assert.Greater(t, response.Usage.TotalTokens, 0, "Total tokens should be greater than 0")
+					assert.Equal(t, response.Usage.PromptTokens+response.Usage.CompletionTokens,
+						response.Usage.TotalTokens, "Total tokens should equal prompt + completion tokens")
+				}
+
 				responseCount++
 
 				// Safety check to prevent infinite loops in case of issues
@@ -110,19 +127,37 @@ func RunChatCompletionStreamTest(t *testing.T, client *bifrost.Bifrost, ctx cont
 		}
 
 	streamComplete:
+		// Validate that we received usage information at some point in the stream
+		assert.True(t, hasReceivedUsage, "Should have received token usage information during streaming")
+
 		// Validate that the last response contains usage information and/or finish reason
 		// with empty choices (typical final chunk pattern)
 		if lastResponse != nil && lastResponse.BifrostResponse != nil {
 			// Check if this is a final metadata chunk (empty choices with usage/finish info)
 			if len(lastResponse.Choices) == 0 && lastResponse.Usage != nil {
+				// Comprehensive validation of final usage
+				assert.Greater(t, lastResponse.Usage.PromptTokens, 0, "Final chunk should have prompt token count")
+				assert.Greater(t, lastResponse.Usage.CompletionTokens, 0, "Final chunk should have completion token count")
 				assert.Greater(t, lastResponse.Usage.TotalTokens, 0, "Final chunk should have total token count")
-				t.Logf("📊 Final metadata chunk - Total tokens: %d", lastResponse.Usage.TotalTokens)
+				assert.Equal(t, lastResponse.Usage.PromptTokens+lastResponse.Usage.CompletionTokens,
+					lastResponse.Usage.TotalTokens, "Total tokens should equal prompt + completion tokens")
+				t.Logf("📊 Final metadata chunk - Prompt: %d, Completion: %d, Total: %d",
+					lastResponse.Usage.PromptTokens,
+					lastResponse.Usage.CompletionTokens,
+					lastResponse.Usage.TotalTokens)
 			} else if len(lastResponse.Choices) > 0 {
 				// Check if final choice has finish reason
 				finalChoice := lastResponse.Choices[0]
 				if finalChoice.FinishReason != nil {
 					t.Logf("🏁 Stream ended with finish reason: %s", *finalChoice.FinishReason)
 				}
+
+				// Even with choices, we should have usage info in the last response or earlier
+				if lastResponse.Usage != nil {
+					assert.Greater(t, lastResponse.Usage.PromptTokens, 0, "Should have prompt tokens")
+					assert.Greater(t, lastResponse.Usage.CompletionTokens, 0, "Should have completion tokens")
+					assert.Greater(t, lastResponse.Usage.TotalTokens, 0, "Should have total tokens")
+				}
 			} else {
 				t.Fatal("Last response should have choices or usage")
 			}
@@ -143,6 +178,9 @@ func RunChatCompletionStreamTest(t *testing.T, client *bifrost.Bifrost, ctx cont
 			} else {
 				// This is a metadata-only chunk, which is valid for final chunks
 				assert.NotNil(t, lastResponse.Usage, "Usage should not be nil")
+				// Additional validation for the usage in metadata-only chunk
+				assert.Greater(t, lastResponse.Usage.PromptTokens, 0, "Metadata chunk should have prompt tokens")
+				assert.Greater(t, lastResponse.Usage.CompletionTokens, 0, "Metadata chunk should have completion tokens")
 			}
 		}
 
@@ -176,6 +214,8 @@ func RunChatCompletionStreamTest(t *testing.T, client *bifrost.Bifrost, ctx cont
 
 			var toolCallDetected bool
 			var responseCount int
+			var hasReceivedUsageWithTools bool
+			var lastResponseWithTools *schemas.BifrostStream
 
 			streamCtx, cancel := context.WithTimeout(ctx, 30*time.Second)
 			defer cancel()
@@ -190,8 +230,25 @@ func RunChatCompletionStreamTest(t *testing.T, client *bifrost.Bifrost, ctx cont
 					}
 
 					require.NotNil(t, response, "Streaming response should not be nil")
+					lastResponseWithTools = response
 					responseCount++
 
+					// Check for usage information in tool call streaming
+					if response.Usage != nil {
+						hasReceivedUsageWithTools = true
+						t.Logf("📊 Tool stream usage - Prompt: %d, Completion: %d, Total: %d",
+							response.Usage.PromptTokens,
+							response.Usage.CompletionTokens,
+							response.Usage.TotalTokens)
+
+						// Validate token counts for tool calls
+						assert.Greater(t, response.Usage.PromptTokens, 0, "Tool stream should have prompt tokens")
+						assert.Greater(t, response.Usage.CompletionTokens, 0, "Tool stream should have completion tokens")
+						assert.Greater(t, response.Usage.TotalTokens, 0, "Tool stream should have total tokens")
+						assert.Equal(t, response.Usage.PromptTokens+response.Usage.CompletionTokens,
+							response.Usage.TotalTokens, "Total should equal prompt + completion for tool stream")
+					}
+
 					for _, choice := range response.Choices {
 						if choice.BifrostStreamResponseChoice != nil {
 							delta := choice.BifrostStreamResponseChoice.Delta
@@ -225,6 +282,15 @@ func RunChatCompletionStreamTest(t *testing.T, client *bifrost.Bifrost, ctx cont
 		toolStreamComplete:
 			assert.Greater(t, responseCount, 0, "Should receive at least one streaming response")
 			assert.True(t, toolCallDetected, "Should detect tool calls in streaming response")
+			assert.True(t, hasReceivedUsageWithTools, "Should have received token usage for tool call stream")
+
+			// Validate final response has proper usage information
+			if lastResponseWithTools != nil && lastResponseWithTools.Usage != nil {
+				assert.Greater(t, lastResponseWithTools.Usage.PromptTokens, 0, "Final tool stream should have prompt tokens")
+				assert.Greater(t, lastResponseWithTools.Usage.CompletionTokens, 0, "Final tool stream should have completion tokens")
+				assert.Greater(t, lastResponseWithTools.Usage.TotalTokens, 0, "Final tool stream should have total tokens")
+			}
+
 			t.Logf("✅ Streaming with tools test completed successfully")
 		})
 	}
diff --git a/transports/bifrost-http/handlers/logging.go b/transports/bifrost-http/handlers/logging.go
index 915a35497b..7029382b9c 100644
--- a/transports/bifrost-http/handlers/logging.go
+++ b/transports/bifrost-http/handlers/logging.go
@@ -147,8 +147,7 @@ func (h *LoggingHandler) getLogs(ctx *fasthttp.RequestCtx) {
 		h.logger.Error("failed to search logs: %v", err)
 		SendError(ctx, fasthttp.StatusInternalServerError, fmt.Sprintf("Search failed: %v", err), h.logger)
 		return
-	}
-
+	}	
 	SendJSON(ctx, result, h.logger)
 }
 
diff --git a/transports/changelog.md b/transports/changelog.md
index 9c26654f07..fe8d5b5f88 100644
--- a/transports/changelog.md
+++ b/transports/changelog.md
@@ -1,4 +1,6 @@
 <!-- The pattern we follow here is to keep the changelog for the latest version -->
 <!-- Old changelogs are automatically attached to the GitHub releases -->
 
-- Fixes pricing computation for nested model names i.e. groq/openai/gpt-oss-20b.
\ No newline at end of file
+- Fix: Users can now delete custom providers from the UI
+- Fix: Token count no longer displays as N/A in certain streaming response cases
+- Fix: Streaming responses now properly display errors on the UI instead of getting stuck in processing state
\ No newline at end of file
diff --git a/transports/version b/transports/version
index 9728bd69ac..9a83513ab0 100644
--- a/transports/version
+++ b/transports/version
@@ -1 +1 @@
-1.2.21
+1.2.22
diff --git a/ui/app/logs/views/logDetailsSheet.tsx b/ui/app/logs/views/logDetailsSheet.tsx
index 45477d305f..52906a3f19 100644
--- a/ui/app/logs/views/logDetailsSheet.tsx
+++ b/ui/app/logs/views/logDetailsSheet.tsx
@@ -274,7 +274,7 @@ export function LogDetailSheet({ log, open, onOpenChange }: LogDetailSheetProps)
 
 				{log.status !== "processing" && (
 					<>
-						{log.output_message && (
+						{log.output_message && !log.error_details?.error.message && (
 							<>
 								<div className="mt-4 flex w-full items-center gap-2">
 									<div className="text-sm font-medium">Response</div>