FreeLLMAPI v0.10.0 brings a unified provider dashboard, monthly key budgets, clearer quota forecasts, and fixes that help requests reach a working model sooner.
Providers and keys
- Manage providers and their models from one dashboard, including custom endpoints and model tests. The Test button respects monthly budgets and records usage. (#1180)
- Edit saved credentials and allowed models without recreating a key. Cloudflare account IDs are preserved when credentials are updated. (#1163)
- Copy working API examples directly from the unified key page. (#1052)
- Added CLōD, Speechify, and BlazeAPI adapters. (#1209)
Budgets and quota outlook
- Set monthly request and token budgets per provider key. Usage survives request-log cleanup, and in-flight requests reserve budget before calling a provider. (#1159)
- See recent request rates, estimated quota exhaustion, and warnings in Quota signals. Forecasts account for the relevant pool and the provider's reset time. (#1157, #1223)
- Recent traffic remains visible when a provider does not report its remaining quota. The dashboard explains missing, stale, low-confidence, and insufficient-data forecasts. (#1227)
- Duration-style reset headers are parsed correctly. Empty probes no longer make old balances look fresh, and a migration repairs retained quota snapshots where observation history is available. (#1227)
Routing and compatibility
- Suspended accounts are temporarily removed from routing across the affected key, instead of retrying every model behind it. Routes that consume most of the failover budget without producing a first byte receive a temporary cooldown. (#1201, #1183)
- Provider-reported request sizes help skip models whose context window is too small, including the output allowance when a provider reports only input size. (#1173)
- Gemini and Ollama compatibility routes fail over when a response contains only reasoning and no usable answer. (#1191)
- Fusion aliases work with saved keys and the Responses API, with streaming and response translation fixes. (#1088)
- Recognize more missing-model errors and record routing decisions consistently across API surfaces. (#1143, #1203)
Dashboard and desktop
- Restoring a model setting to its catalog value clears its Local Override badge. (#1188)
- Depleted models are dimmed and move below usable models in score mode. Manual ordering is preserved. (#1190)
- Improved Windows and Linux desktop contrast and removed unnecessary macOS window padding. (#1177)
- Added the Chinese recurring-credit gateway guide and refreshed documentation. (#1186)
Updating
Download the installer for your platform below, or use ghcr.io/tashfeenahmed/freellmapi:v0.10.0 for Docker. The latest Docker tag also points to this release.
Existing keys and profiles are retained. Database migrations run automatically at startup. Back up your data before updating a self-hosted instance.
Quota forecasts still depend on recent traffic and reliable provider-reported balances. An unavailable forecast does not mean your quota is exhausted.
⭐ Like the free router? Go Premium, the live signed catalog, $19/yr, cancel anytime.