@blockrun/clawrouter 0.12.44 → 0.12.46

package/dist/index.js

@@ -3157,63 +3157,74 @@ var DEFAULT_ROUTING_CONFIG = {
     confidenceThreshold: 0.7
   },
   // Auto (balanced) tier configs - current default smart routing
+  // Benchmark-tuned 2026-03-16: latency-ranked via blockrun.ai x402 end-to-end
   tiers: {
     SIMPLE: {
-      primary: "moonshot/kimi-k2.5",
-      // $0.60/$3.00 - best quality/price for simple tasks
+      primary: "xai/grok-4-fast-non-reasoning",
+      // 1,143ms, $0.20/$0.50 — fastest overall
       fallback: [
+        "xai/grok-3-mini",
+        // 1,202ms, $0.30/$0.50
         "google/gemini-2.5-flash",
-        // 60% retention (best), fast growth (+800%)
+        // 1,238ms, 60% retention (best)
         "google/gemini-2.5-flash-lite",
-        // 1M context, ultra cheap ($0.10/$0.40)
+        // 1,353ms, 1M context, ultra cheap ($0.10/$0.40)
         "deepseek/deepseek-chat",
-        // 41% retention
+        // 1,431ms, 41% retention
         "nvidia/gpt-oss-120b"
-        // FREE fallback
+        // 1,252ms, FREE fallback
       ]
     },
     MEDIUM: {
-      primary: "moonshot/kimi-k2.5",
-      // $0.50/$2.40 - strong tool use, proper function call format
+      primary: "xai/grok-4-1-fast-non-reasoning",
+      // 1,244ms, $0.20/$0.50 — fast + tool calling
       fallback: [
         "deepseek/deepseek-chat",
-        // 41% retention
+        // 1,431ms, 41% retention
+        "moonshot/kimi-k2.5",
+        // 1,646ms, strong tool use quality
         "google/gemini-2.5-flash",
-        // 60% retention, cheap fast model
+        // 1,238ms, 60% retention
         "google/gemini-2.5-flash-lite",
-        // 1M context, ultra cheap ($0.10/$0.40)
-        "xai/grok-4-1-fast-non-reasoning"
-        // Upgraded Grok 4.1
+        // 1,353ms, 1M context ($0.10/$0.40)
+        "xai/grok-3-mini"
+        // 1,202ms, $0.30/$0.50
       ]
     },
     COMPLEX: {
       primary: "google/gemini-3.1-pro",
-      // Newest Gemini 3.1 - upgraded from 3.0
+      // 1,609ms — fast flagship quality
       fallback: [
         "google/gemini-2.5-flash",
-        // 60% retention, cheap failsafe before expensive models
+        // 1,238ms, cheap failsafe before expensive models
         "google/gemini-2.5-flash-lite",
-        // CRITICAL: 1M context, ultra-cheap failsafe ($0.10/$0.40)
+        // 1,353ms, 1M context, ultra-cheap failsafe ($0.10/$0.40)
         "google/gemini-3-pro-preview",
-        // 3.0 fallback
+        // 1,352ms
         "google/gemini-2.5-pro",
-        "deepseek/deepseek-chat",
+        // 1,294ms
         "xai/grok-4-0709",
-        "openai/gpt-5.4",
-        // Newest flagship, same price as 4o
-        "openai/gpt-4o",
-        "anthropic/claude-sonnet-4.6"
+        // 1,348ms
+        "deepseek/deepseek-chat",
+        // 1,431ms
+        "anthropic/claude-sonnet-4.6",
+        // 2,110ms — quality fallback
+        "openai/gpt-5.4"
+        // 6,213ms — slowest but highest quality
       ]
     },
     REASONING: {
       primary: "xai/grok-4-1-fast-reasoning",
-      // Upgraded Grok 4.1 reasoning $0.20/$0.50
+      // 1,454ms, $0.20/$0.50
       fallback: [
+        "xai/grok-4-fast-reasoning",
+        // 1,298ms, $0.20/$0.50
         "deepseek/deepseek-reasoner",
-        // Cheap reasoning model
+        // 1,454ms, cheap reasoning
         "openai/o4-mini",
-        // Newer and cheaper than o3 ($1.10 vs $2.00)
+        // 2,328ms ($1.10/$4.40)
         "openai/o3"
+        // 2,862ms
       ]
     }
   },
@@ -3221,27 +3232,30 @@ var DEFAULT_ROUTING_CONFIG = {
   ecoTiers: {
     SIMPLE: {
       primary: "nvidia/gpt-oss-120b",
-      // FREE! $0.00/$0.00
+      // 1,252ms, FREE! $0.00/$0.00
       fallback: [
         "google/gemini-2.5-flash-lite",
-        "google/gemini-2.5-flash",
-        "deepseek/deepseek-chat"
+        // 1,353ms, $0.10/$0.40
+        "xai/grok-4-fast-non-reasoning",
+        // 1,143ms, $0.20/$0.50
+        "google/gemini-2.5-flash"
+        // 1,238ms
       ]
     },
     MEDIUM: {
       primary: "google/gemini-2.5-flash-lite",
-      // $0.10/$0.40 - cheapest capable with 1M context
-      fallback: ["google/gemini-2.5-flash", "deepseek/deepseek-chat", "nvidia/gpt-oss-120b"]
+      // 1,353ms, $0.10/$0.40 - cheapest capable with 1M context
+      fallback: ["xai/grok-4-fast-non-reasoning", "google/gemini-2.5-flash", "deepseek/deepseek-chat", "nvidia/gpt-oss-120b"]
     },
     COMPLEX: {
       primary: "google/gemini-2.5-flash-lite",
-      // $0.10/$0.40 - 1M context handles complexity
-      fallback: ["google/gemini-2.5-flash", "deepseek/deepseek-chat", "xai/grok-4-0709"]
+      // 1,353ms, $0.10/$0.40 - 1M context handles complexity
+      fallback: ["xai/grok-4-0709", "google/gemini-2.5-flash", "deepseek/deepseek-chat"]
     },
     REASONING: {
       primary: "xai/grok-4-1-fast-reasoning",
-      // $0.20/$0.50
-      fallback: ["deepseek/deepseek-reasoner"]
+      // 1,454ms, $0.20/$0.50
+      fallback: ["xai/grok-4-fast-reasoning", "deepseek/deepseek-reasoner"]
     }
   },
   // Premium tier configs - best quality (blockrun/premium)
@@ -3287,57 +3301,71 @@ var DEFAULT_ROUTING_CONFIG = {
     },
     REASONING: {
       primary: "anthropic/claude-sonnet-4.6",
-      // $3/$15 - best for reasoning/instructions
+      // 2,110ms, $3/$15 - best for reasoning/instructions
       fallback: [
         "anthropic/claude-opus-4.6",
-        "anthropic/claude-opus-4.6",
+        // 2,139ms
+        "xai/grok-4-1-fast-reasoning",
+        // 1,454ms, cheap fast reasoning
         "openai/o4-mini",
-        // Newer and cheaper than o3 ($1.10 vs $2.00)
-        "openai/o3",
-        "xai/grok-4-1-fast-reasoning"
+        // 2,328ms ($1.10/$4.40)
+        "openai/o3"
+        // 2,862ms
       ]
     }
   },
   // Agentic tier configs - models that excel at multi-step autonomous tasks
   agenticTiers: {
     SIMPLE: {
-      primary: "moonshot/kimi-k2.5",
-      // Cheaper than Haiku ($0.5/$2.4 vs $1/$5), larger context
+      primary: "xai/grok-4-1-fast-non-reasoning",
+      // 1,244ms, $0.20/$0.50 — fast tool calling
       fallback: [
-        "anthropic/claude-haiku-4.5",
-        "xai/grok-4-1-fast-non-reasoning",
-        "openai/gpt-4o-mini"
+        "openai/gpt-4o-mini",
+        // 2,764ms, $0.15/$0.60 - reliable tool compliance
+        "moonshot/kimi-k2.5",
+        // 1,646ms, strong tool use quality
+        "anthropic/claude-haiku-4.5"
+        // 2,305ms
       ]
     },
     MEDIUM: {
       primary: "moonshot/kimi-k2.5",
-      // $0.50/$2.40 - strong tool use, handles function calls correctly
+      // 1,646ms, $0.60/$3.00 - strong tool use, proper function calls
       fallback: [
+        "xai/grok-4-1-fast-non-reasoning",
+        // 1,244ms, fast fallback
+        "openai/gpt-4o-mini",
+        // 2,764ms, reliable tool calling
         "anthropic/claude-haiku-4.5",
-        "deepseek/deepseek-chat",
-        "xai/grok-4-1-fast-non-reasoning"
+        // 2,305ms
+        "deepseek/deepseek-chat"
+        // 1,431ms
       ]
     },
     COMPLEX: {
       primary: "anthropic/claude-sonnet-4.6",
+      // 2,110ms — best agentic quality
       fallback: [
         "anthropic/claude-opus-4.6",
-        // Latest Opus - best agentic
-        "openai/gpt-5.4",
-        // Newest flagship
+        // 2,139ms — top quality
         "google/gemini-3.1-pro",
-        // Newest Gemini
-        "google/gemini-3-pro-preview",
-        "xai/grok-4-0709"
+        // 1,609ms
+        "xai/grok-4-0709",
+        // 1,348ms
+        "openai/gpt-5.4"
+        // 6,213ms — slow but highest quality fallback
       ]
     },
     REASONING: {
       primary: "anthropic/claude-sonnet-4.6",
-      // Strong tool use + reasoning for agentic tasks
+      // 2,110ms — strong tool use + reasoning
       fallback: [
         "anthropic/claude-opus-4.6",
+        // 2,139ms
         "xai/grok-4-1-fast-reasoning",
+        // 1,454ms
         "deepseek/deepseek-reasoner"
+        // 1,454ms
       ]
     }
   },
@@ -7415,17 +7443,9 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
             hasTools
           });
           if (hasTools && routingDecision.tier === "SIMPLE") {
-            const simpleRoutingDecision = route(prompt, systemPrompt, maxTokens, {
-              ...routerOpts,
-              routingProfile: routingProfile ?? void 0,
-              hasTools: false
-            });
-            if (simpleRoutingDecision.tier === "SIMPLE") {
-              console.log(
-                `[ClawRouter] SIMPLE+tools: using non-agentic model ${simpleRoutingDecision.model} (tools present but query is trivial)`
-              );
-              routingDecision = simpleRoutingDecision;
-            }
+            console.log(
+              `[ClawRouter] SIMPLE+tools: keeping agentic model ${routingDecision.model} (tools need reliable function-call support)`
+            );
           }
           if (existingSession) {
             const tierRank = {
@@ -7742,7 +7762,7 @@ async function proxyRequest(req, res, apiBase, payFetch, options, routerOpts, de
     } else {
       modelsToTry = modelId ? [modelId] : [];
     }
-    if (!modelsToTry.includes(FREE_MODEL)) {
+    if (!hasTools && !modelsToTry.includes(FREE_MODEL)) {
       modelsToTry.push(FREE_MODEL);
     }
     let upstream;