Models & context

Models and credits: choose the right answer and understand its cost

Search the model catalog, set response length, estimate credit use, add supported provider keys, and handle credit limits.

Before you start

The model picker is available in the starting panel, each Canvas draft, and the Chat composer. It lists models available to your account. Featured offers a curated shortlist; All models by provider lets you browse the catalog, and the search field can match a model, provider, or identifier. Labels such as Top, Value, and Free help compare available choices; they are not guarantees about a particular answer.

When to use this

Choose a model based on the task, the context it must read, and the cost shown in the app. Use a stronger model for a difficult synthesis if appropriate; try a Free model for routine questions when its daily availability suits you. Select a response length based on how much detail you need.

Steps

Open the model picker beside the prompt or node’s model name. Search for a provider or model, or browse Featured and All models by provider.
Select a model. If your prompt includes images, choose a model with vision; incompatible models are disabled while images are attached. A model marked Free has no credit cost, but can be temporarily busy or subject to daily limits.
Before generating, choose Short, Medium, or Extended from the length control. New nodes start at Short unless you changed the default in Settings → General. The nearby estimate is for that model and selected length. The available maximum may depend on your plan, the model, or whether you use an eligible personal provider key.
Generate the answer. Its actual credit charge appears on the answer or Canvas node. The charge includes the answer and may include context or web searches, so it can differ from an estimate.
To set the default for new nodes, open Settings → General and choose a value under Default response length. You can still change length on an individual node or message.
To use your own supported provider key, open Settings → API keys, choose OpenAI, Anthropic, or Google, and connect or replace the key. A paid plan is required for BYOK. Never paste a provider key into a prompt or a connected-app tool call.
Check your current usage from the account menu. For current plan limits and features, open neurothread.ai/pricing rather than relying on an old saved number.

Worked example

For “Summarize these meeting notes and list decisions, owners, and unresolved questions,” choose a model that can read the attached document, select Medium, and inspect the estimate before sending. If you only need a checklist, switch to Short. After the response, use the displayed charge to understand actual usage. If you need to compare styles, create a separate branch with another model rather than replacing the first result.

What happens next

Answers retain the model that produced them. A new branch normally inherits the model from the answer it continues, and you can select another model before generating. If an answer reaches its length limit, Continue creates a new node for the same model to pick up where it stopped. For another model’s take on an existing answer, use Other model in Chat or change the model in a Canvas draft.

Cost and limitations

Credit quotas, pricing, and plan features can change; use the account’s usage display and the live pricing page. Monthly credits do not roll over. Free models still have daily/shared availability limits and may be busy. Your own OpenAI, Anthropic, or Google key is billed by that provider, not by Neurothread credits, and provider billing rules apply. BYOK is not available for every provider or model. Web search is separate from the model call: it can use Neurothread credits while credits remain, including when BYOK is active, and may be unavailable after those credits run out. Longer responses and attached context can affect the model cost. An answer stopped mid-generation keeps its partial text and is billed for what was produced.

Troubleshooting

No model matches: clear or broaden the search; try a provider name or model identifier.
A model is disabled: it may not support attached images, may be temporarily unavailable, or may be locked by your plan for the current mode.
Insufficient credits: try a shorter length or lower-cost model, select an available Free model, add an eligible key, or review plan options.
Provider rejected your key: check that the key belongs to the selected provider, is active, and has provider-side billing access; reconnect it in Settings → API keys.
Estimate differs from the charge: the estimate is not a fixed price; actual answer length, context, and searches influence the final amount.