FreeLLMAPI v0.13.3 extends the output-limit fix to every provider and makes custom endpoints findable on the Keys page.
Routing
- Output limits are learned from every provider, not just Groq and Ollama Cloud. v0.13.2 taught the router to learn a model's output limit when it turns down a large
max_tokens(as Claude Code sends), but it only recognised the wording Groq and Ollama Cloud use. OpenRouter, Together, Fireworks, DeepInfra, Mistral and custom endpoints phrase the same refusal differently, so their models were still set aside and marked as unable to use tools. The router now recognises those phrasings too, retries within the limit, and keeps the model in rotation. (#1367)
Dashboard
- Search on the Keys page matches a custom endpoint's URL. A custom provider row shows its endpoint host, and searching for that host now finds it, the same way the fallback table's search already did. (#1368)
Updating
Download the installer for your platform below, or use ghcr.io/tashfeenahmed/freellmapi:v0.13.3 for Docker. The latest Docker tag also points to this release. Desktop users on v0.13.2 will see this update in the Updates panel.
Existing keys and profiles are retained. This release has no database migrations.
⭐ Like the free router? Go Premium, the live signed catalog, $19/yr, cancel anytime.