diff --git a/core/providers/utils/utils.go b/core/providers/utils/utils.go
index 3d722bc71b..ab33f6e77c 100644
--- a/core/providers/utils/utils.go
+++ b/core/providers/utils/utils.go
@@ -1055,12 +1055,31 @@ func CheckContextAndGetRequestBody(ctx context.Context, request RequestBodyGette
 				}
 			}
 		}
+		// Drop unsupported parameters identified by the litellmcompat plugin
+		jsonBody = dropUnsupportedParams(ctx, jsonBody)
 		return jsonBody, nil
 	} else {
 		return rawBody, nil
 	}
 }
 
+// dropUnsupportedParams removes top-level JSON fields listed in the
+// BifrostContextKeyLiteLLMCompatDroppedParams context key. The drop list is
+// computed by the litellmcompat plugin by comparing the request's Params
+// against the model's supported parameters allowlist from the catalog.
+func dropUnsupportedParams(ctx context.Context, jsonBody []byte) []byte {
+	droppedParams, ok := ctx.Value(schemas.BifrostContextKeyLiteLLMCompatDroppedParams).([]string)
+	if !ok || len(droppedParams) == 0 {
+		return jsonBody
+	}
+	for _, param := range droppedParams {
+		if modified, err := sjson.DeleteBytes(jsonBody, param); err == nil {
+			jsonBody = modified
+		}
+	}
+	return jsonBody
+}
+
 // SetExtraHeadersHTTP sets additional headers from NetworkConfig to the standard HTTP request.
 // This allows users to configure custom headers for their provider requests.
 // Header keys are canonicalized using textproto.CanonicalMIMEHeaderKey to avoid duplicates.
diff --git a/core/schemas/bifrost.go b/core/schemas/bifrost.go
index c2a64e28fb..8ac9137a9e 100644
--- a/core/schemas/bifrost.go
+++ b/core/schemas/bifrost.go
@@ -255,6 +255,7 @@ const (
 	BifrostContextKeySessionID                           BifrostContextKey = "bifrost-session-id"                         // string session ID for the request (session stickiness)
 	BifrostContextKeySessionTTL                          BifrostContextKey = "bifrost-session-ttl"                        // time.Duration session TTL for the request (session stickiness)
 	BifrostContextKeyMCPExtraHeaders                     BifrostContextKey = "bifrost-mcp-extra-headers"                  // map[string][]string (these headers are forwarded only to the MCP while tool execution if they are in the allowlist of the MCP client)
+	BifrostContextKeyLiteLLMCompatDroppedParams          BifrostContextKey = "litellmcompat-dropped-params"               // []string (set by litellmcompat plugin - parameter names to drop from the provider request body)
 )
 
 const (
@@ -964,4 +965,4 @@ type BifrostErrorExtraFields struct {
 	RawResponse    interface{}   `json:"raw_response,omitempty"`
 	LiteLLMCompat  bool          `json:"litellm_compat,omitempty"`
 	KeyStatuses    []KeyStatus   `json:"key_statuses,omitempty"`
-}
\ No newline at end of file
+}
diff --git a/framework/modelcatalog/main.go b/framework/modelcatalog/main.go
index 35730395ad..c40e4776b4 100644
--- a/framework/modelcatalog/main.go
+++ b/framework/modelcatalog/main.go
@@ -55,6 +55,10 @@ type ModelCatalog struct {
 	// Values are normalized output types: "chat_completion", "responses", "text_completion"
 	supportedOutputs map[string][]string
 
+	// Pre-parsed supported parameters index (keyed by model name, populated from model parameters supported_parameters)
+	// Values are parameter names the model accepts (e.g., "temperature", "top_p", "tools")
+	supportedParams map[string][]string
+
 	// Background sync worker
 	syncTicker *time.Ticker
 	done       chan struct{}
@@ -209,6 +213,7 @@ func Init(ctx context.Context, config *Config, configStore configstore.ConfigSto
 		unfilteredModelPool:    make(map[schemas.ModelProvider][]string),
 		baseModelIndex:         make(map[string]string),
 		supportedOutputs:       make(map[string][]string),
+		supportedParams:        make(map[string][]string),
 		done:                   make(chan struct{}),
 		shouldSyncPricingFunc:  shouldSyncPricingFunc,
 		distributedLockManager: configstore.NewDistributedLockManager(configStore, logger, configstore.WithDefaultTTL(30*time.Second)),
@@ -907,36 +912,123 @@ func (mc *ModelCatalog) IsResponsesSupported(model string, provider schemas.Mode
 	return ok && slices.Contains(outputs, "responses")
 }
 
-// buildSupportedOutputsIndex parses supported_endpoints from model parameters data
-// and rebuilds the supportedOutputs index with normalized output type names.
+// GetSupportedParameters returns the list of supported parameter names for a model.
+// Returns nil if the model is not found in the catalog.
+func (mc *ModelCatalog) GetSupportedParameters(model string) []string {
+	mc.mu.RLock()
+	params, ok := mc.supportedParams[model]
+	mc.mu.RUnlock()
+	if !ok {
+		return nil
+	}
+	// Return a copy to prevent external modification
+	result := make([]string, len(params))
+	copy(result, params)
+	return result
+}
+
+// buildSupportedOutputsIndex parses supported_endpoints and model parameters/capabilities
+// from model parameters data and rebuilds the in-memory indexes.
 func (mc *ModelCatalog) buildSupportedOutputsIndex(paramsData map[string]json.RawMessage) {
-	newIndex := make(map[string][]string, len(paramsData))
+	newOutputsIndex := make(map[string][]string, len(paramsData))
+	newParamsIndex := make(map[string][]string, len(paramsData))
 
 	for model, data := range paramsData {
-		var params struct {
-			SupportedEndpoints []string `json:"supported_endpoints"`
-		}
-		if err := json.Unmarshal(data, &params); err != nil || len(params.SupportedEndpoints) == 0 {
+		var parsed modelParametersParseResult
+		if err := json.Unmarshal(data, &parsed); err != nil {
 			continue
 		}
-		outputs := make([]string, 0, len(params.SupportedEndpoints))
-		for _, endpoint := range params.SupportedEndpoints {
-			if normalized := normalizeEndpointToOutputType(endpoint); normalized != "" {
-				if !slices.Contains(outputs, normalized) {
-					outputs = append(outputs, normalized)
+
+		// Build supported outputs from endpoints
+		if len(parsed.SupportedEndpoints) > 0 {
+			outputs := make([]string, 0, len(parsed.SupportedEndpoints))
+			for _, endpoint := range parsed.SupportedEndpoints {
+				if normalized := normalizeEndpointToOutputType(endpoint); normalized != "" {
+					if !slices.Contains(outputs, normalized) {
+						outputs = append(outputs, normalized)
+					}
 				}
 			}
+			if len(outputs) > 0 {
+				newOutputsIndex[model] = outputs
+			}
 		}
-		if len(outputs) > 0 {
-			newIndex[model] = outputs
+
+		// Build supported params from model_parameters IDs and supports_* flags
+		supported := extractSupportedParams(&parsed)
+		if len(supported) > 0 {
+			newParamsIndex[model] = supported
 		}
 	}
 
 	mc.mu.Lock()
-	mc.supportedOutputs = newIndex
+	mc.supportedOutputs = newOutputsIndex
+	mc.supportedParams = newParamsIndex
 	mc.mu.Unlock()
 }
 
+// modelParametersParseResult is the parsed result type used by buildSupportedOutputsIndex.
+type modelParametersParseResult struct {
+	SupportedEndpoints []string `json:"supported_endpoints"`
+	ModelParameters    []struct {
+		ID string `json:"id"`
+	} `json:"model_parameters"`
+	SupportsFunctionCalling         *bool `json:"supports_function_calling"`
+	SupportsParallelFunctionCalling *bool `json:"supports_parallel_function_calling"`
+	SupportsToolChoice              *bool `json:"supports_tool_choice"`
+	SupportsReasoning               *bool `json:"supports_reasoning"`
+	SupportsServiceTier             *bool `json:"supports_service_tier"`
+	SupportsPromptCaching           *bool `json:"supports_prompt_caching"`
+}
+
+// extractSupportedParams builds a list of supported OpenAI-compatible parameter
+// names from model_parameters[].id values and supports_* boolean flags.
+func extractSupportedParams(parsed *modelParametersParseResult) []string {
+	var supported []string
+	addParam := func(name string) {
+		if !slices.Contains(supported, name) {
+			supported = append(supported, name)
+		}
+	}
+
+	// From model_parameters[].id — map IDs to request param names
+	for _, mp := range parsed.ModelParameters {
+		switch mp.ID {
+		case "reasoning_effort", "reasoning_summary":
+			addParam("reasoning")
+		case "web_search":
+			addParam("web_search_options")
+		case "promptTools", "image_detail", "stream":
+			// skip — not top-level request parameters
+		default:
+			addParam(mp.ID)
+		}
+	}
+
+	// From supports_* boolean flags
+	if parsed.SupportsFunctionCalling != nil && *parsed.SupportsFunctionCalling {
+		addParam("tools")
+	}
+	if parsed.SupportsParallelFunctionCalling != nil && *parsed.SupportsParallelFunctionCalling {
+		addParam("parallel_tool_calls")
+	}
+	if parsed.SupportsToolChoice != nil && *parsed.SupportsToolChoice {
+		addParam("tool_choice")
+	}
+	if parsed.SupportsReasoning != nil && *parsed.SupportsReasoning {
+		addParam("reasoning")
+	}
+	if parsed.SupportsServiceTier != nil && *parsed.SupportsServiceTier {
+		addParam("service_tier")
+	}
+	if parsed.SupportsPromptCaching != nil && *parsed.SupportsPromptCaching {
+		addParam("prompt_cache_key")
+		addParam("prompt_cache_retention")
+	}
+
+	return supported
+}
+
 // populateModelPool populates the model pool with all available models per provider (thread-safe)
 func (mc *ModelCatalog) populateModelPoolFromPricingData() {
 	// Acquire write lock for the entire rebuild operation
@@ -1019,6 +1111,7 @@ func NewTestCatalog(baseModelIndex map[string]string) *ModelCatalog {
 		baseModelIndex:      baseModelIndex,
 		pricingData:         make(map[string]configstoreTables.TableModelPricing),
 		supportedOutputs:    make(map[string][]string),
+		supportedParams:     make(map[string][]string),
 		done:                make(chan struct{}),
 	}
 }
diff --git a/plugins/litellmcompat/dropparams.go b/plugins/litellmcompat/dropparams.go
new file mode 100644
index 0000000000..9a7169d57c
--- /dev/null
+++ b/plugins/litellmcompat/dropparams.go
@@ -0,0 +1,175 @@
+package litellmcompat
+
+import (
+	"slices"
+
+	"github.com/maximhq/bifrost/core/schemas"
+)
+
+// computeUnsupportedParams checks each parameter field on the request's Params
+// and returns the JSON field names of parameters that are set but not in the
+// model's supported parameters allowlist. It does NOT mutate the request.
+func computeUnsupportedParams(req *schemas.BifrostRequest, supportedParams []string) []string {
+	if req == nil {
+		return nil
+	}
+	switch {
+	case req.ChatRequest != nil && req.ChatRequest.Params != nil:
+		return unsupportedChatParams(req.ChatRequest.Params, supportedParams)
+	case req.ResponsesRequest != nil && req.ResponsesRequest.Params != nil:
+		return unsupportedResponsesParams(req.ResponsesRequest.Params, supportedParams)
+	case req.TextCompletionRequest != nil && req.TextCompletionRequest.Params != nil:
+		return unsupportedTextCompletionParams(req.TextCompletionRequest.Params, supportedParams)
+	}
+	return nil
+}
+
+func unsupportedChatParams(p *schemas.ChatParameters, supported []string) []string {
+	var dropped []string
+	if p.Audio != nil && !slices.Contains(supported, "audio") {
+		dropped = append(dropped, "audio")
+	}
+	if p.FrequencyPenalty != nil && !slices.Contains(supported, "frequency_penalty") {
+		dropped = append(dropped, "frequency_penalty")
+	}
+	if p.LogitBias != nil && !slices.Contains(supported, "logit_bias") {
+		dropped = append(dropped, "logit_bias")
+	}
+	if p.LogProbs != nil && !slices.Contains(supported, "logprobs") {
+		dropped = append(dropped, "logprobs")
+	}
+	if p.MaxCompletionTokens != nil && !slices.Contains(supported, "max_completion_tokens") {
+		dropped = append(dropped, "max_completion_tokens")
+	}
+	if p.Metadata != nil && !slices.Contains(supported, "metadata") {
+		dropped = append(dropped, "metadata")
+	}
+	if p.ParallelToolCalls != nil && !slices.Contains(supported, "parallel_tool_calls") {
+		dropped = append(dropped, "parallel_tool_calls")
+	}
+	if p.Prediction != nil && !slices.Contains(supported, "prediction") {
+		dropped = append(dropped, "prediction")
+	}
+	if p.PresencePenalty != nil && !slices.Contains(supported, "presence_penalty") {
+		dropped = append(dropped, "presence_penalty")
+	}
+	if p.PromptCacheKey != nil && !slices.Contains(supported, "prompt_cache_key") {
+		dropped = append(dropped, "prompt_cache_key")
+	}
+	if p.PromptCacheRetention != nil && !slices.Contains(supported, "prompt_cache_retention") {
+		dropped = append(dropped, "prompt_cache_retention")
+	}
+	if p.Reasoning != nil && !slices.Contains(supported, "reasoning") {
+		dropped = append(dropped, "reasoning")
+	}
+	if p.ResponseFormat != nil && !slices.Contains(supported, "response_format") {
+		dropped = append(dropped, "response_format")
+	}
+	if p.Seed != nil && !slices.Contains(supported, "seed") {
+		dropped = append(dropped, "seed")
+	}
+	if p.ServiceTier != nil && !slices.Contains(supported, "service_tier") {
+		dropped = append(dropped, "service_tier")
+	}
+	if len(p.Stop) > 0 && !slices.Contains(supported, "stop") {
+		dropped = append(dropped, "stop")
+	}
+	if p.Temperature != nil && !slices.Contains(supported, "temperature") {
+		dropped = append(dropped, "temperature")
+	}
+	if p.TopLogProbs != nil && !slices.Contains(supported, "top_logprobs") {
+		dropped = append(dropped, "top_logprobs")
+	}
+	if p.TopP != nil && !slices.Contains(supported, "top_p") {
+		dropped = append(dropped, "top_p")
+	}
+	if p.ToolChoice != nil && !slices.Contains(supported, "tool_choice") {
+		dropped = append(dropped, "tool_choice")
+	}
+	if len(p.Tools) > 0 && !slices.Contains(supported, "tools") {
+		dropped = append(dropped, "tools")
+	}
+	if p.Verbosity != nil && !slices.Contains(supported, "verbosity") {
+		dropped = append(dropped, "verbosity")
+	}
+	return dropped
+}
+
+func unsupportedResponsesParams(p *schemas.ResponsesParameters, supported []string) []string {
+	var dropped []string
+	if p.MaxOutputTokens != nil && !slices.Contains(supported, "max_output_tokens") {
+		dropped = append(dropped, "max_output_tokens")
+	}
+	if p.MaxToolCalls != nil && !slices.Contains(supported, "max_tool_calls") {
+		dropped = append(dropped, "max_tool_calls")
+	}
+	if p.Metadata != nil && !slices.Contains(supported, "metadata") {
+		dropped = append(dropped, "metadata")
+	}
+	if p.ParallelToolCalls != nil && !slices.Contains(supported, "parallel_tool_calls") {
+		dropped = append(dropped, "parallel_tool_calls")
+	}
+	if p.PromptCacheKey != nil && !slices.Contains(supported, "prompt_cache_key") {
+		dropped = append(dropped, "prompt_cache_key")
+	}
+	if p.Reasoning != nil && !slices.Contains(supported, "reasoning") {
+		dropped = append(dropped, "reasoning")
+	}
+	if p.ServiceTier != nil && !slices.Contains(supported, "service_tier") {
+		dropped = append(dropped, "service_tier")
+	}
+	if p.Temperature != nil && !slices.Contains(supported, "temperature") {
+		dropped = append(dropped, "temperature")
+	}
+	if p.Text != nil && !slices.Contains(supported, "text") {
+		dropped = append(dropped, "text")
+	}
+	if p.TopLogProbs != nil && !slices.Contains(supported, "top_logprobs") {
+		dropped = append(dropped, "top_logprobs")
+	}
+	if p.TopP != nil && !slices.Contains(supported, "top_p") {
+		dropped = append(dropped, "top_p")
+	}
+	if p.ToolChoice != nil && !slices.Contains(supported, "tool_choice") {
+		dropped = append(dropped, "tool_choice")
+	}
+	if len(p.Tools) > 0 && !slices.Contains(supported, "tools") {
+		dropped = append(dropped, "tools")
+	}
+	return dropped
+}
+
+func unsupportedTextCompletionParams(p *schemas.TextCompletionParameters, supported []string) []string {
+	var dropped []string
+	if p.FrequencyPenalty != nil && !slices.Contains(supported, "frequency_penalty") {
+		dropped = append(dropped, "frequency_penalty")
+	}
+	if p.LogitBias != nil && !slices.Contains(supported, "logit_bias") {
+		dropped = append(dropped, "logit_bias")
+	}
+	if p.LogProbs != nil && !slices.Contains(supported, "logprobs") {
+		dropped = append(dropped, "logprobs")
+	}
+	if p.MaxTokens != nil && !slices.Contains(supported, "max_tokens") {
+		dropped = append(dropped, "max_tokens")
+	}
+	if p.N != nil && !slices.Contains(supported, "n") {
+		dropped = append(dropped, "n")
+	}
+	if p.PresencePenalty != nil && !slices.Contains(supported, "presence_penalty") {
+		dropped = append(dropped, "presence_penalty")
+	}
+	if p.Seed != nil && !slices.Contains(supported, "seed") {
+		dropped = append(dropped, "seed")
+	}
+	if len(p.Stop) > 0 && !slices.Contains(supported, "stop") {
+		dropped = append(dropped, "stop")
+	}
+	if p.Temperature != nil && !slices.Contains(supported, "temperature") {
+		dropped = append(dropped, "temperature")
+	}
+	if p.TopP != nil && !slices.Contains(supported, "top_p") {
+		dropped = append(dropped, "top_p")
+	}
+	return dropped
+}
diff --git a/plugins/litellmcompat/main.go b/plugins/litellmcompat/main.go
index 59de6fedb8..5e9414ed5b 100644
--- a/plugins/litellmcompat/main.go
+++ b/plugins/litellmcompat/main.go
@@ -86,6 +86,20 @@ func (p *LiteLLMCompatPlugin) PreLLMHook(ctx *schemas.BifrostContext, req *schem
 	// Apply request transforms in sequence
 	req = transformTextToChatRequest(ctx, req, p.modelCatalog, p.logger)
 	req = transformChatToResponsesRequest(ctx, req, p.modelCatalog, p.logger)
+
+	// Compute unsupported parameters to drop based on model catalog allowlist
+	if ctx != nil && p.modelCatalog != nil {
+		model := getModelFromRequest(req)
+		if model != "" {
+			if supportedParams := p.modelCatalog.GetSupportedParameters(model); supportedParams != nil {
+				droppedParams := computeUnsupportedParams(req, supportedParams)
+				if len(droppedParams) > 0 {
+					ctx.SetValue(schemas.BifrostContextKeyLiteLLMCompatDroppedParams, droppedParams)
+				}
+			}
+		}
+	}
+
 	return req, nil, nil
 }
 
@@ -103,7 +117,25 @@ func (p *LiteLLMCompatPlugin) PostLLMHook(ctx *schemas.BifrostContext, result *s
 	return result, bifrostErr, nil
 }
 
+// getModelFromRequest extracts the model name from a BifrostRequest,
+// checking each request type in order.
+func getModelFromRequest(req *schemas.BifrostRequest) string {
+	if req == nil {
+		return ""
+	}
+	if req.ChatRequest != nil {
+		return req.ChatRequest.Model
+	}
+	if req.ResponsesRequest != nil {
+		return req.ResponsesRequest.Model
+	}
+	if req.TextCompletionRequest != nil {
+		return req.TextCompletionRequest.Model
+	}
+	return ""
+}
+
 // Cleanup performs plugin cleanup
 func (p *LiteLLMCompatPlugin) Cleanup() error {
 	return nil
-}
+}
\ No newline at end of file