Major Change to Free Tier Providers

[mystyc]

Has there been a major change to free tier providers?
I am getting all sorts of errors about "free tiers" lacking sufficient funds, models being retired that I used last week, quota limits on providers I have barely used, etc...
Cheaper Inference, Kimi, OpenCode, and even Antigravity act like they either don't have a free tier anymore, or the tier was gimped.

• GitHub [400]: The requested model is not supported. Why are some models gh/ or github/. Haven't they already ended?

• Cheaper Inference 402, insufficient balance (on a free tier) and 404 not available

• No Kimi free tier? [403]: Your current subscription does not have access to Kimi Code right now. Upgrade your plan to keep coding with Kimi Code: https://www.kimi.com/code/#pricing

• nvidia 410: The model 'openai/gpt-oss-120b' has reached its end of life on 2026-09-03T08:00:00Z and is no longer available.

• [429]: Antigravity upstream error (429): Resource has been exhausted (e.g. check quota).

• auto/claude-sonnet [502] dva/claude-sonnet-5-low: provider — Devin CLI not found: devin. Install the official Devin CLI or set CLI_DEVIN_AGENTIC_BIN. (HTTP 502) and "Devin CLI is not installed or can't find sandbox absolute path." It isn't installed because I don't use the Devin CLI.
  And yet all these provider connections are healthy?

I'm looking at provider sites, but they seem evasive when talking about their free tiers.
Is the era of free tier stacks over?

[diegosouzapw]

Hi @mystyc. Most of these errors come from the providers themselves, and "healthy" in OmniRoute only means the connection's credentials work, not that the account still has free quota for a given model. Going through your list:

• GitHub 400 The requested model is not supported. gh/ is just the alias of the GitHub Copilot provider (id github), so gh/... and github/... are the same thing. That 400 comes from Copilot when the model isn't enabled on your Copilot plan or has been dropped from it.
• Cheaper Inference 402 insufficient balance. That provider has no free tier in OmniRoute's catalog. It's prepaid, so 402 means the account has no balance.
• Kimi 403 ... does not have access to Kimi Code. Kimi Code is tied to a Kimi subscription, and the message says your current plan doesn't include it. OmniRoute can't change that.
• NVIDIA 410 ... gpt-oss-120b has reached its end of life. NVIDIA retired the model, but our NVIDIA catalog still lists it, so that one is on us. I've opened fix(nvidia): catalog still lists openai/gpt-oss-120b after NVIDIA retired it (410) #18418 to remove it.
• Antigravity 429 Resource has been exhausted. Google's quota for that account ran out. It resets on Google's schedule.
• auto/claude-sonnet -> dva/claude-sonnet-5-low, Devin CLI not found. This one is a bug in OmniRoute. The Devin CLI bridge is supposed to stay out of auto/* until the CLI is installed and ready, but the family pools (auto/claude-sonnet, auto/glm, ...) skip that check. I've opened fix(autoCombo): auto/<family> pools skip the local-CLI readiness gate and route to Devin CLI when it is not installed #18417 for it. Workaround: open Providers -> Devin CLI Agentic Bridge and switch it off. Disabled no-auth providers are left out of every auto pool, including the family ones.

So free tiers haven't ended across the board, but they do change often, and providers tighten them without much notice. The Free Tiers and Free Provider Rankings pages in the dashboard show what OmniRoute knows about each one. Spreading requests over several providers in one combo is what keeps a free setup working when one of them runs dry.