Troubleshooting

Atlas error: Provider has no models (ProviderNoModelsError)

Updated 7 min read

Atlas raises ProviderNoModelsError because the provider authenticated and loaded, but its model map came back empty, typically because the models endpoint returned nothing or the account has no entitled models. Fix it by running `atlas models` to confirm the provider section is genuinely empty, re-authenticating in case an expired token is yielding an empty list rather than a hard 401, checking the provider dashboard that your account or plan actually grants model access, and removing any model pin from atlas.json so Atlas can fetch the live list.

Why does Atlas say a provider has no models

Atlas says a provider has no models because Provider.defaultModel raises NoModelsError when the provider resolves but its model map is empty. The credential worked and the provider loaded. The model list simply came back with 0 entries for that providerID.

ProviderNoModelsError sits one step past authentication. The provider resolved, which means Atlas found a credential and constructed the client without throwing. What failed is the model map: it is empty. Two situations produce an empty map. The provider's models endpoint returned nothing, or the account behind the credential has no entitled models. Both look identical from inside Atlas, because Atlas can only report what the provider hands back. The empty model picker for that provider is the visible half of the same fact. Everything downstream, including selecting a default model, is blocked until that map has at least one entry.

How do I fix ProviderNoModelsError in Atlas

Step 1 of the documented fix is to run `atlas models` and confirm the provider section is genuinely empty. Confirming the emptiness before changing anything keeps you from chasing a plan or billing problem that does not exist, because a stale picker is not the same as an empty model map.

Work the fix in order. Run `atlas models` first and read the section for the providerID named in the error. If models are listed there, the problem is elsewhere. If the section is genuinely empty, re-authenticate that provider next, because an expired token often yields an empty model list rather than a hard 401 and this is the case people miss most often. If re-authentication does not repopulate the list, check the provider dashboard to confirm your account or plan actually grants model access. Finally, if you pinned models in atlas.json, remove the pin so Atlas can fetch the live list.

Can an expired token cause an empty model list in Atlas

Yes. An expired token often yields an empty model list rather than a hard 401, which is why re-authenticating is step two of the ProviderNoModelsError fix. Atlas sees a provider that loaded and a model map with zero entries, so it raises NoModelsError instead of an auth error.

The behavior is counterintuitive, and it is worth stating plainly: a token that has expired can produce a silent empty list rather than a clean rejection. From the Atlas side, that is indistinguishable from an account with no entitled models. Re-authenticate the provider named in the error before you conclude anything about your plan or your billing. If the model map repopulates after re-authentication, the token was the whole story. If it stays empty after a fresh credential, the token was not the problem and you should move on to the provider dashboard and to any model pin in atlas.json.

How to check whether your account is entitled to models

Step 3 of the documented fix is to check the provider dashboard directly and confirm your account or plan actually grants model access. Atlas raises ProviderNoModelsError whenever the model map is empty, and an account with 0 entitled models produces exactly that, even though the credential itself is completely valid.

Entitlement and authentication are separate. A key can be real, current, and correctly stored in Atlas, and still map to an account whose plan grants access to zero models. The provider dashboard is the only authoritative place to settle that question, because Atlas can only report the model map it receives. Look at the plan tied to the exact key you configured, not to another key on the same account. If the dashboard shows no model access, no amount of re-authenticating or reconfiguring inside Atlas will populate the model map, and the fix is on the provider side rather than in Atlas.

Does a model pin in atlas.json cause ProviderNoModelsError

A model pin in atlas.json can cause ProviderNoModelsError, which is why removing the pin is step 4 of the documented fix. Remove the pin so Atlas can fetch the live model list for the provider instead of being held to a static set that may no longer resolve.

Pinning models in atlas.json is useful when you want a fixed, predictable set. The cost is that the pinned list, not the provider's live list, is what Atlas works from. If those pinned entries do not resolve against the provider, the model map ends up empty and Provider.defaultModel raises NoModelsError for that providerID. Removing the pin lets Atlas fetch the live list again, which both fixes the immediate error and tells you what the provider genuinely offers. Re-add a pin afterwards if you want one, using a model name you have now seen in the live list.

How to verify the Provider has no models fix worked

Run `atlas models` to verify the fix. When the provider section that listed 0 models now lists models, the model map is populated, Provider.defaultModel has something to select, and ProviderNoModelsError will not fire again for that providerID. The Atlas model picker fills in for that provider at the same time.

`atlas models` is the direct read of the condition behind the error, which makes it the right verification command. Look specifically at the section for the providerID that appeared in the ProviderNoModelsError message, not at the output as a whole. Other providers listing models tells you nothing about this one. If the section is populated, you are done. If it is still empty after re-authenticating, checking entitlements on the provider dashboard, and removing any model pin in atlas.json, the provider is genuinely returning nothing and the remaining work is with the provider, not with Atlas.

How to fix it

  1. 01Run `atlas models` and confirm the provider section is genuinely empty.
  2. 02Re-authenticate the provider: an expired token often yields an empty model list rather than a hard 401.
  3. 03Check the provider dashboard that your account or plan actually grants model access.
  4. 04If you pinned models in atlas.json, remove the pin so Atlas can fetch the live list.

Frequently asked questions

how to fix ProviderNoModelsError in atlas
Run `atlas models` and confirm the provider section is genuinely empty, re-authenticate the provider, check the provider dashboard that your plan grants model access, and remove any model pin from atlas.json so Atlas can fetch the live list.
why is my atlas model picker empty for one provider
An empty Atlas model picker for one provider means that provider resolved but its model map is empty. Provider.defaultModel raises NoModelsError in that case, usually because the models endpoint returned nothing or the account has no entitled models.
atlas provider authenticated but no models listed
Authentication succeeding while no models are listed is the classic ProviderNoModelsError case. Re-authenticate first, because an expired token often yields an empty model list rather than a hard 401, then check entitlements on the provider dashboard.
can an expired api key show an empty model list instead of a 401
Yes. In Atlas, an expired token often yields an empty model list rather than a hard 401. That empty model map is what makes Provider.defaultModel raise NoModelsError, so re-authenticating is the first thing to try.
should I remove pinned models from atlas.json
If you pinned models in atlas.json and hit ProviderNoModelsError, remove the pin so Atlas can fetch the live model list from the provider. You can re-add a pin later using a model name you have seen in the live list.
how do I know if my plan grants model access
Check the provider dashboard for the exact key you configured in Atlas. A valid credential attached to an account with no entitled models produces an empty model map, and Atlas raises ProviderNoModelsError for that providerID.
how do I confirm the atlas no models error is fixed
Run `atlas models` and look at the section for the providerID from the error. If that section now lists models, the model map is populated and ProviderNoModelsError will not fire again for that provider.

Try Atlas in your terminal

The terminal-native AI coding agent. Free core, single binary.

Install Atlas

Related guides

Atlas with Claude Opus 4.6: Setup, Cost, and Tradeoffs in 2026

Claude Opus 4.6 gives Atlas a 1M token context at $5 per Mtok input, $25 per Mtok output. Setup steps, cost math, and when to pick a newer Opus instead.

Atlas with Claude Opus 4.7 in 2026: 1M Context at $5 / $25 per Mtok

Claude Opus 4.7 drives Atlas with a 1M tokens (1,000,000) window, 128K max output, and $5 per Mtok input, $25 per Mtok output. Setup, tradeoffs, and when to move on.

Atlas with Kimi K2 0905: 262,144 Tokens In and Out, 2026

Run Atlas on Kimi K2 0905 in 2026. Moonshot's September refresh gives 256K tokens (262,144) context and output at $0.60 per Mtok in, $2.50 per Mtok out.

Atlas with Qwen3 30B-A3B (Ollama): the MoE throughput trade in 2026

Qwen3 30B-A3B (Ollama) in Atlas: 30B total parameters, roughly 3B active per token, 19GB of weights, 256K tokens (262,144) of context, Free (self-hosted).

Atlas with Qwen3.5 Plus: A Million-Token Window for $0.40 per Mtok in 2026

Qwen3.5 Plus gives Atlas a 1M tokens (1,000,000) context window at $0.40 per Mtok input and $2.40 per Mtok output. What the million tokens buy, and what closed weights cost.

Atlas with Claude Opus 4.5: The Price Break Opus in 2026

Claude Opus 4.5 runs Atlas on a 200K context at $5 per Mtok input, $25 per Mtok output, down from $15/$75 on Opus 4.1. Setup, extended thinking, and tradeoffs.

Atlas with GPT-5.1 Codex mini: Wide Subagent Fan Out in 2026

GPT-5.1 Codex mini in Atlas: an OpenAI fast tier at $0.25 per Mtok input, $2 per Mtok output, built for parallel subagents, bulk sweeps, and cheap summaries.

Atlas with Command R 35B (Ollama): A RAG-Native Model for Retrieval-Heavy Work in 2026

Run Atlas on Command R 35B (Ollama): Cohere's 19GB RAG and tool-use model with a 128K context, free self-hosted. Check the research license before commercial use.

Browse this resource hub