Error codes

Every error from the BitsRun API, the dashboard, the command line and the apps comes with the same three things:

{ "error": { "code": "model_starting", "message": "Your model is starting...", "requestId": "01k9a2b3c4d5e6f7g8h9j0k1m2" } }

• code never changes, so your code can check it.
• message explains it in plain words.
• requestId identifies the request. Include it when you write to support@bitsrun.com.

The command line prints the message and the request id. Add --json to see the full error.

Codes by area

Calling your endpoint: unauthorized, not_found, payment_required, rate_limited, model_starting, model_unavailable, model_preparing, no_capacity, context_too_long, bad_request

Deploying a model: invalid_model, model_not_found, model_gated, model_no_access, hf_token_invalid, model_importing, imports_busy, import_failed, model_custom_code, model_task_unsupported, model_choose_file, model_split_unsupported, model_quant_not_found, model_too_large, gpu_too_small, invalid_gpu, invalid_name, invalid_idle_timeout, endpoint_limit_reached, credit_required

Endpoint settings: endpoint_preparing, always_on_stop, always_on_credit, priority_unavailable, preset_unavailable, invalid_context, invalid_parallel, too_many_changes

Uploads and storage: invalid_upload_file, upload_not_ready, upload_not_open, upload_incomplete, upload_not_gguf, upload_invalid_model, upload_in_use, storage_quota_reached

Account and billing: signin_required, reauth_required, invalid_code, limit_reached, invalid_amount, payments_unavailable, invalid_plan, already_subscribed, no_subscription, demo_read_only, terms_required, terms_outdated, adult_required, name_invalid

Adapters: adapter_not_found, adapter_loading, invalid_adapter_source, adapter_files, adapter_config_invalid, adapter_unsupported, adapter_wrong_base, adapter_rank_too_high, adapter_modules, adapter_too_large, adapter_file_invalid, adapter_limit_reached, adapter_pool_full, adapter_in_use, invalid_adapters, adapter_endpoint_unsupported, upload_is_adapter

General: unavailable, not_implemented, internal

unauthorized

HTTP status 401.

What happened: The request had no API key, a mistyped key, or a key that was revoked. It also appears for an account that was deleted.

How to fix it: Send the header Authorization: Bearer brk_... with a key from the API keys page. If the key was revoked or lost, create a new one. See API keys.

not_found

HTTP status 404.

What happened: The address or endpoint does not exist for this key. This also covers deleted endpoints, endpoints in another account, paths the endpoint does not serve, and keys limited to a different endpoint. All of these answer the same way on purpose.

How to fix it: Copy the endpoint URL again from its card or with bitsrun status <endpoint>. Check the path (for example /v1/chat/completions) and that the key works with this endpoint.

payment_required

HTTP status 402.

What happened: Your balance is $0 or below, so endpoints are stopped. The demo account always gets this code.

How to fix it: Add credit on the Billing page. Endpoints answer again within seconds. Set a balance email or monthly top-up so it does not happen again. See Add credit and auto top-up.

rate_limited

HTTP status 429.

What happened: Your account already has as many requests open at once as your plan allows (10, 50 or 150).

How to fix it: Retry after a short pause that grows each time (1, 2, then 4 seconds). Send fewer requests in parallel, or move to a bigger plan. See Limits.

model_starting

HTTP status 200 inside a response that waited, or 503 with the X-BitsRun-Wait: 0 header or when 4 of your requests already wait on that endpoint.

What happened: The endpoint was asleep and is still starting. Requests wait up to 8 minutes for it; this answer means it took longer (a very large model or a slow download), and the error arrives in the body because the response had already started. With X-BitsRun-Wait: 0 you get it as a 503 after about 45 seconds, with a Retry-After: 30 header.

How to fix it: Send the request again; the model keeps starting meanwhile. Start-up time is never billed. See Cold starts.

model_unavailable

HTTP status 503.

What happened: The model failed to start several times in a row, so BitsRun stopped waking it.

How to fix it: Open the endpoint in the dashboard and check its status and settings. Try a different performance preset or a larger GPU, or write to support with the request id.

model_preparing

HTTP status 503.

What happened: The endpoint is still copying its model into your storage, so there is nothing to start yet.

How to fix it: Wait a few minutes; the card shows the copy's progress. The endpoint starts on the first request after the copy is ready.

no_capacity

HTTP status 503.

What happened: Every GPU of this size was busy for 30 minutes, so BitsRun stopped trying for now. You were not billed while it waited. The Retry-After header says when it tries again.

How to fix it: Try again in a few minutes, or choose a different GPU size in the endpoint's settings. Priority (on L4 and L40S) starts on a dedicated GPU right away. See Sleep, always-on and priority.

context_too_long

HTTP status 400.

What happened: The request is longer than the endpoint can read at once (its context, or memory length).

How to fix it: Shorten the conversation or start a new one, split long documents, or raise Memory length under Performance.

bad_request

HTTP status 400.

What happened: The request body is not valid JSON, or a value is out of range.

How to fix it: Check that the body is valid JSON with the fields the path expects. See Chat and completions.

invalid_model

HTTP status 400.

What happened: The model name is not in a form BitsRun understands.

How to fix it: Use an Ollama name (qwen3:8b), a Hugging Face repo or link (hf:unsloth/Qwen3-8B-GGUF, or paste the page or file link), or one of your uploads (upload:<id>). See Choose a model.

model_not_found

HTTP status 400.

What happened: We could not find that model on Ollama or Hugging Face.

How to fix it: Check the spelling and the tag (qwen3:8b), or the Hugging Face repo name. For a private repo, add a Hugging Face token that can read it. See Gated and private repos.

model_gated

HTTP status 400.

What happened: The Hugging Face repo is gated: its author asks you to accept their terms first.

How to fix it: Accept the terms on the repo page at huggingface.co, add your Hugging Face token on the API keys page (or run bitsrun hf-token set), then deploy again.

model_no_access

HTTP status 403.

What happened: Your Hugging Face account does not have access to this repo yet.

How to fix it: Accept the terms on the repo page at huggingface.co. Some authors review requests first, so wait for their approval, then deploy again.

hf_token_invalid

HTTP status 400.

What happened: Hugging Face did not accept the token you pasted.

How to fix it: Create a read token at huggingface.co/settings/tokens and paste it again.

model_importing

HTTP status 409.

What happened: We are copying this model into your private storage. Large models take a few minutes.

How to fix it: Deploy again when the copy is ready. You can follow it under Uploads. The command line waits for you.

imports_busy

HTTP status 409.

What happened: You already have 2 models copying into your storage.

How to fix it: Wait for one to finish (the Endpoints page shows progress), then try again.

import_failed

HTTP status 502.

What happened: We could not copy this model from Hugging Face.

How to fix it: Deploy it again to retry, or use bitsrun retry <endpoint> for an endpoint whose copy failed. If it keeps failing, write to support with the request id.

model_custom_code

HTTP status 400.

What happened: The model needs its own Python code to run (trust_remote_code). BitsRun never runs code that comes with a model.

How to fix it: Pick a version of the model that works without custom code, or a GGUF version. See Models that cannot be deployed.

model_task_unsupported

HTTP status 400.

What happened: The repo is a kind of model BitsRun does not serve yet, such as vision, audio or reranking. Today it serves text generation and embedding models.

How to fix it: Pick a text generation or embedding model.

model_choose_file

HTTP status 400.

What happened: The repo keeps its model files in folders, so BitsRun cannot tell which one you want.

How to fix it: Open the .gguf file you want on Hugging Face and paste its link instead.

model_split_unsupported

HTTP status 400.

What happened: The repo only has GGUF files split into several parts, which BitsRun cannot serve yet.

How to fix it: Pick a repo that offers single-file quants.

model_quant_not_found

HTTP status 400.

What happened: The repo has no file with the quantization you named.

How to fix it: Leave the quant empty to use Q4_K_M, or name one the repo lists, for example hf:owner/repo:Q8_0. See Quantization.

model_too_large

HTTP status 400.

What happened: The model is larger than the biggest GPU size BitsRun offers today.

How to fix it: Pick a smaller model or a more compressed version (a lower quant). See The model sizer.

gpu_too_small

HTTP status 400.

What happened: The model needs more GPU memory than the size you chose.

How to fix it: Pick a larger GPU, or leave the GPU empty and BitsRun picks the smallest one that fits.

invalid_gpu

HTTP status 400.

What happened: The GPU size is not one BitsRun offers.

How to fix it: Use l4, l40s or 4xl4, or leave it empty.

invalid_name

HTTP status 400.

What happened: The name has characters that are not allowed, or is too long.

How to fix it: Use 1 to 64 letters, numbers, spaces, dots, dashes or underscores.

invalid_idle_timeout

HTTP status 400.

What happened: The sleep setting is outside the allowed range.

How to fix it: Use a whole number of seconds from 60 to 3600 (1 minute to 1 hour). The command line also accepts 90s, 5m or 1h.

endpoint_limit_reached

HTTP status 409.

What happened: Your plan has no room for another endpoint. Endpoints still copying their model count too.

How to fix it: Delete an endpoint you no longer use, or move to a bigger plan. See Plans.

credit_required

HTTP status 402.

What happened: Uploading or copying a model needs a positive balance, because stored models are kept for you.

How to fix it: Add credit, then try again.

endpoint_preparing

HTTP status 409.

What happened: The endpoint is still copying its model, so its settings cannot change yet.

How to fix it: Wait until the copy is ready, then change the setting.

always_on_stop

HTTP status 409.

What happened: The endpoint is always on, so stopping it would start it again right away.

How to fix it: Turn off Always on in its settings first, then stop it.

always_on_credit

HTTP status 402.

What happened: Always-on needs at least one day of credit at its rate for this GPU size.

How to fix it: Add credit, then turn always-on on again. See How billing works.

priority_unavailable

HTTP status 409.

What happened: Priority is not offered for this GPU size (4x L4).

How to fix it: Leave priority off for this endpoint.

preset_unavailable

HTTP status 409.

What happened: That performance option does not apply to this model.

How to fix it: Pick one of the options the Performance panel shows.

invalid_context

HTTP status 400.

What happened: The memory length is outside what this model and GPU allow.

How to fix it: Choose a memory length between 2,048 tokens and the most the panel offers for this GPU.

invalid_parallel

HTTP status 400.

What happened: Requests at once must be between 1 and 8.

How to fix it: Choose a number from 1 to 8.

too_many_changes

HTTP status 429.

What happened: This endpoint changed settings 6 times in the last hour.

How to fix it: Wait a little, then try again.

invalid_upload_file

HTTP status 400.

What happened: The files are not a single .gguf file or a model folder, or a file is smaller than 1 MB.

How to fix it: Upload one .gguf file, or a folder with config.json and .safetensors weights. See Upload your own model.

upload_not_ready

HTTP status 400.

What happened: That upload is not finished yet.

How to fix it: Finish uploading it, then deploy it.

upload_not_open

HTTP status 409.

What happened: This upload is already finished, so it cannot take more parts.

How to fix it: Start a new upload to send another file.

upload_incomplete

HTTP status 400.

What happened: Some parts of the file did not arrive.

How to fix it: Run the upload again. bitsrun upload retries parts for you.

upload_not_gguf

HTTP status 400.

What happened: The file is not a valid GGUF model, so it was removed.

How to fix it: Check that it is a .gguf file exported by llama.cpp or a similar tool.

upload_invalid_model

HTTP status 400.

What happened: The files are not a model BitsRun can serve, so they were removed.

How to fix it: A model folder needs config.json and valid .safetensors weights, as saved by Hugging Face transformers or sentence-transformers.

upload_in_use

HTTP status 409.

What happened: An endpoint still serves this upload.

How to fix it: Delete that endpoint first, then delete the upload.

storage_quota_reached

HTTP status 409.

What happened: This model would go past your plan's most storage.

How to fix it: Delete a model you no longer use, or move to a bigger plan. See Model storage.

signin_required

HTTP status 401.

What happened: Your dashboard, app or command line session ended or was signed out.

How to fix it: Sign in again. On the command line, run bitsrun login.

reauth_required

HTTP status 403.

What happened: Deleting your account needs a sign-in from the last 10 minutes.

How to fix it: Confirm it is you with a new email code, then try again.

invalid_code

HTTP status 400.

What happened: The command line sign-in code is not valid or has expired (codes last 5 minutes and work once).

How to fix it: Run bitsrun login again.

limit_reached

HTTP status 409.

What happened: You reached an account limit, such as 20 API keys or 20 command line sessions.

How to fix it: Delete one you no longer use, or write to support@bitsrun.com to raise it.

invalid_amount

HTTP status 400.

What happened: Top-ups must be whole dollars from $10 to $10,000.

How to fix it: Pick an amount in that range.

payments_unavailable

HTTP status 503.

What happened: Adding credit is briefly unavailable.

How to fix it: Try again in a few minutes, or write to support@bitsrun.com.

invalid_plan

HTTP status 400.

What happened: The plan name is not one BitsRun offers.

How to fix it: Choose Pro or Business.

already_subscribed

HTTP status 409.

What happened: You already have a plan.

How to fix it: Use the manage link on the Billing page to change or cancel it.

no_subscription

HTTP status 409.

What happened: You do not have a plan to manage yet.

How to fix it: Subscribe from the Billing page first.

demo_read_only

HTTP status 403.

What happened: This is the demo account, so changes are turned off.

How to fix it: Sign up with your own email to deploy models and add credit.

terms_required

HTTP status 403.

What happened: You have not accepted the current Terms and Acceptable Use Policy yet. Viewing and deleting still work; other changes wait.

How to fix it: Open the dashboard, review the terms on Before you start, and accept them.

terms_outdated

HTTP status 409.

What happened: Our terms changed while the page was open, so the version you accepted is not the current one.

How to fix it: Reload the page, read the current version and accept it.

adult_required

HTTP status 400.

What happened: BitsRun is for people 18 or older, and the age box was not ticked.

How to fix it: Tick the box to confirm you are 18 or older, then continue.

name_invalid

HTTP status 400.

What happened: The first or last name has characters we do not accept or is too long.

How to fix it: Use letters, spaces, hyphens, apostrophes or periods, up to 64 characters each. Leave the last name empty if you have only one name.

adapter_not_found

HTTP status 404.

What happened: The model field named an adapter id (lora-...) that is not yours, is not ready yet, or is not added to this endpoint.

How to fix it: Check the id on the Adapters page. An adapter that is still copying shows its progress there. Add it to the endpoint in its settings.

adapter_loading

HTTP status 503.

What happened: Your adapter is still being copied onto the machine that serves the endpoint.

How to fix it: Send the request again in about 15 seconds.

invalid_adapter_source

HTTP status 400.

What happened: The adapter source is not a Hugging Face repo or one of your uploads.

How to fix it: Enter hf:owner/repo, a Hugging Face link, or upload:<id>.

adapter_files

HTTP status 400.

What happened: The repo or upload does not have adapter_config.json and adapter_model.safetensors at the top. Adapters saved only as .bin files are not accepted.

How to fix it: Save the adapter again with safetensors (PEFT's save_pretrained does this by default), then register it again.

adapter_config_invalid

HTTP status 400.

What happened: adapter_config.json is not valid JSON or is missing peft_type, base_model_name_or_path, r, lora_alpha or target_modules.

How to fix it: Save the adapter again with PEFT, then register it again.

adapter_unsupported

HTTP status 400.

What happened: The adapter is not a plain LoRA (for example DoRA, prompt tuning, modules_to_save, bias training or per-layer settings).

How to fix it: Train a plain LoRA, then register it again.

adapter_wrong_base

HTTP status 400.

What happened: The adapter was trained on a model we do not serve adapters for. Today that is Qwen3 8B only.

How to fix it: Train on Qwen/Qwen3-8B (or a copy such as unsloth/Qwen3-8B).

adapter_rank_too_high

HTTP status 400.

What happened: The adapter's rank (r) is above 64, the most we serve.

How to fix it: Train it again with r set to 64 or less.

adapter_modules

HTTP status 400.

What happened: The adapter changes layers we cannot serve.

How to fix it: Set target_modules to names from q_proj, k_proj, v_proj, o_proj, gate_proj, up_proj and down_proj.

adapter_too_large

HTTP status 400.

What happened: The adapter is larger than 1 GB, the most we accept.

How to fix it: Use a rank of 64 or less on the attention and MLP projections; that always fits.

adapter_file_invalid

HTTP status 400.

What happened: adapter_model.safetensors does not match its adapter_config.json, so we removed it.

How to fix it: Save the adapter again with PEFT, then register it again.

adapter_limit_reached

HTTP status 409.

What happened: Your account holds as many adapters as it can.

How to fix it: Delete one you no longer use, or write to support@bitsrun.com to raise the limit.

adapter_pool_full

HTTP status 409.

What happened: We cannot take more adapters for this model right now.

How to fix it: Try again later, or write to support@bitsrun.com and we will make room.

adapter_in_use

HTTP status 409.

What happened: One of your endpoints uses this adapter.

How to fix it: Remove it from that endpoint's settings first, then delete it.

invalid_adapters

HTTP status 400.

What happened: The endpoint's adapter list is not valid.

How to fix it: List up to 4 of your adapter ids, each once, or an empty list to remove them.

adapter_endpoint_unsupported

HTTP status 400.

What happened: Adapters work only on endpoints that run Qwen3 8B with vLLM, deployed from hf:Qwen/Qwen3-8B or hf:Qwen/Qwen3-8B-FP8.

How to fix it: Deploy one of those for your adapters, or remove the adapters from this endpoint first.

upload_is_adapter

HTTP status 400.

What happened: This upload is a LoRA adapter, not a model.

How to fix it: Register it on the Adapters page, add it to your Qwen3 8B endpoint, and call it by its adapter id.

unavailable

HTTP status 503.

What happened: The service is briefly unavailable.

How to fix it: Try again shortly. If it keeps happening, write to support with the request id.

not_implemented

HTTP status 501.

What happened: This feature is not available yet.

How to fix it: Check these docs for what is available today, or write to support if you need it.

internal

HTTP status 500.

What happened: Something went wrong on our side.

How to fix it: Try again. If it happens again, write to support@bitsrun.com with the request id.