Skip to content
SeoFreshUp Docs

OpenRouter errors

OpenRouter sends specific error codes when something goes wrong. Here’s how to decode the most common ones.

401 Unauthorized

OpenRouter: 401 No auth credentials found

Means: your API key is missing, malformed, or revoked.

Fixes

  1. Check that the key is pasted correctly in Settings → API Keys & Models — no leading/trailing spaces, starts with sk-or-
  2. Click 🔌 Test the key — if it fails, the key is invalid
  3. Go to openrouter.ai/keys, revoke the old one, create a new one, paste in the plugin

402 Payment Required

OpenRouter: 402 You have insufficient credits

Means: your OpenRouter balance hit zero.

Fix

  1. Go to openrouter.ai/credits
  2. Top up (recommended: $10-20 for ~6 months of normal usage)
  3. Retry the action — credits update instantly

404 Model Not Found

OpenRouter: 404 X is not a valid model ID

Means: the model you configured doesn’t exist (deprecated or typo).

Fix

  1. Settings → API Keys & Models
  2. For each role (Standard, Deep, Rewrite, Image), check the model ID
  3. Either pick from the curated dropdown, or click 🔍 Discover available models to see live OpenRouter catalog

429 Rate Limit

OpenRouter: 429 Rate limit exceeded

Means: too many requests too fast. Common when you click “Analyse complète” on a 5000-article blog.

Fix

  1. Wait 30 seconds, retry
  2. For bulk operations, the plugin paces requests automatically (max 5 concurrent), but very large blogs (>2000 articles) may still hit limits
  3. Consider switching to a less-loaded model (e.g., from claude-haiku-4.5 to gpt-4o-mini)

500 / 502 / 503 Server Errors

OpenRouter or the upstream model provider had a temporary issue.

Fix

Wait 1-5 minutes and retry. If persistent (>30 min), check status.openrouter.ai.

”Réponse de réécriture non parseable”

The model returned non-JSON output (rare with modern models, common with older ones like Llama 7B).

Fixes

  1. Switch the Rewrite model to something more reliable: anthropic/claude-haiku-4.5 or openai/gpt-4o-mini
  2. Avoid :online variants for rewrite (web search models often produce free-form output)
  3. Avoid models with very small context windows

”Pas de message dans la réponse” / Empty content

The model returned a response but with empty content.

Causes

  • Content moderation kicked in (rare for SEO content but possible if your article mentions sensitive topics)
  • The prompt was too long for the model’s context window — the model truncated and returned empty

Fixes

  1. Try a different model with larger context (Claude Opus 4.7 = 200k tokens, Gemini 2.5 Pro = 2M tokens)
  2. If your article is >5000 words, consider splitting it into sections and rewriting each separately

”Image gen: empty image in response”

The image gen model returned a chat response with no image attached.

Fixes

  1. Check the model supports image generation (must have images modality on OpenRouter)
  2. Switch to a tested model: google/gemini-3-pro-image-preview
  3. Check that “Activer la génération d’image” is on in Settings

Cost monitoring

Track your OpenRouter spending at openrouter.ai/activity. Set a low-balance alert at $1-2 to avoid surprise zero-balance failures during a long bulk job.

What’s next?