fix: traduccion con max_tokens 600 + priorizar nemotron-3-ultra-550b (rate de exito 100%)

2026-08-05 06:19:05 +08:00
parent 636d629034
commit 91ffc714f3
2 changed files with 14 additions and 13 deletions
@@ -27,15 +27,16 @@ export async function translateTitle(title: string): Promise<string | null> {
  // Si ya parece estar en español, no traducir (heurística ligera).
  if (isProbablySpanish(clean)) return clean

-  // Modelos en orden de preferencia (benchmark real 2026-08):
-  // nemotron-3-nano-omni y openrouter/free traducen bien.
+  // Modelos en orden de preferencia (benchmark + tests reales 2026-08):
+  // nemotron-3-ultra-550b traduce bien con max_tokens>=600.
  const models = [
+    "nvidia/nemotron-3-ultra-550b-a55b:free",
    "nvidia/nemotron-3-nano-omni-30b-a3b-reasoning:free",
+    "nemotron-3-ultra-free",
    "openrouter/free",
    "deepseek-v4-flash-free",
    "big-pickle",
    "tencent/hy3:free",
-    "nvidia/nemotron-3-ultra-550b-a55b:free",
    "mimo-v2.5-free",
    "cohere/north-mini-code:free",
    "inclusionai/ling-3.0-flash:free",
@@ -62,7 +63,7 @@ export async function translateTitle(title: string): Promise<string | null> {
            { role: "system", content: SYSTEM_PROMPT },
            { role: "user", content: clean },
          ],
-          max_tokens: 200,
+          max_tokens: 600, // >=500: los reasoning models gastan los primeros tokens en razonamiento
          temperature: 0.3,
        }),
        signal: AbortSignal.timeout(90000),