Error reference

Free LLM API errors, explained and fixed.

Each page quotes the exact error a provider returns, explains what causes it, shows the fix on the provider itself, and then shows how freelm handles it automatically by failing over to another free provider. Limits are third-party facts, so every page carries the date it was checked.

Provider

Google Gemini API

429 RESOURCE_EXHAUSTED

Gemini API 429 RESOURCE_EXHAUSTED: quota exceeded, and how to fix it

Gemini's 429 RESOURCE_EXHAUSTED means one model's quota ran out, usually the free daily cap. How to tell a daily cap from a per-minute burst, and fix each.

400 FAILED_PRECONDITION

Gemini: "User location is not supported for the API use"

Gemini rejects requests from unsupported regions with 400 FAILED_PRECONDITION. What decides the location, the traps (edge runtimes, IPv6), and the fixes.

400 API_KEY_INVALID

Gemini "API key not valid. Please pass a valid API key."

Gemini's 400 API_KEY_INVALID: a mistyped or unloaded key, AI Studio's new AQ. keys in old clients, or an unrestricted key Google now blocks. How to find which.

503 UNAVAILABLE

Gemini 503 "The model is overloaded" or "experiencing high demand"

Gemini's 503 UNAVAILABLE ('The model is overloaded', now 'experiencing high demand') is a temporary capacity limit on one model, not your key. How to handle it.

Provider

OpenRouter

429 free-models-per-day

OpenRouter "Rate limit exceeded: free-models-per-day" (and per-min)

OpenRouter's free models allow 20 requests a minute and 50 a day, or 1,000 after buying 10 credits. How to read the 429, when it resets, and the upstream case.

404 No endpoints found

OpenRouter "No endpoints found that support tool use"

OpenRouter returns 404 'No endpoints found' when no provider for the model supports your request: tools, images, or your privacy and guardrail settings. Fixes.

Provider

Groq

429 rate_limit_exceeded

Groq "Rate limit reached for model ... on tokens per minute (TPM)"

Groq's free plan limits each model per organization: 30 requests a minute, 1K a day, 8K tokens a minute, 200K a day. How to read the 429 and the 413.

Provider

Mistral

429 rate_limited (1300)

Mistral API "Rate limit exceeded" on the free plan

Mistral's 429 on Free mode: per-model requests per second and tokens per minute, or a model capped at zero. How to read the headers and what to change.

Provider

Z.ai (GLM)

429 code 1302

Z.ai GLM "Rate limit reached for requests" (1302) and 1305

Z.ai returns 1302 when you exceed a model's concurrency or rate limit, and 1305 when the model is overloaded. Causes and fixes for the free GLM Flash models.

Provider

Thinking models (Gemini, gpt-oss)

content: null, finish_reason: length

Empty response from Gemini or gpt-oss: content null, finish_reason length

Thinking models spend max_tokens on hidden reasoning and return empty content with finish_reason length. Why Gemini 3 and gpt-oss do it, and the fix for each.