Use models in Dify, n8n and FastGPT

Recommended models available now

Chosen for workflows: cheap, fast and good enough for bulk nodes like classification, extraction and rewriting. Switch only the nodes that need reasoning quality to a stronger model.

Platform prices per 1M tokens, billed on actual usage. The list follows the live pricing API; retired models never show up here. Pick a model and the config and cost below update to match.

Exact configuration

The API base and the selected model qwen3.8-flash are already filled in. Replace the key with your own.

# Dify → Settings → Model Provider → OpenAI-API-compatible → Add model
Model Type:        LLM
Model Name:        qwen3.8-flash
API Key:           sk-your-key
API endpoint URL:  https://token.easyapi.com/v1
Completion mode:   Chat
Model context size: see the model detail page
# enable both "Function calling" and "Stream"

Full Dify integration guide →

Get one request working first

Swap in your key and run it as is. A 200 means the base URL, key and model are all right; otherwise match the error against the list below.

curl https://token.easyapi.com/v1/chat/completions \
  -H "Authorization: Bearer sk-your-key" \
  -H "Content-Type: application/json" \
  -d '{"model":"qwen3.8-flash","messages":[{"role":"user","content":"Hello"}]}'

What to know

• Platform model dropdowns usually do not list custom models. Type the model name in; use the name from the model detail page.
• Dify asks for context size and max output by hand. Oversized values do not error but mislead the platform about available room; copy them from the model detail page.
• Bulk workflow cost is driven by call count, not unit price. The example below assumes 5,000 calls a day; scale by your node count.
• For nodes that need tool calls or structured output, confirm the selected model supports them and enable Function calling in the provider settings.

Common errors

Validation fails when adding the model

The base URL must end with /v1 and the model name is case-sensitive without a vendor prefix. Verify the key and model name with the curl request on this page first.

Node fails with 401 / 403

401 is a wrong or disabled key; 403 means the key's model allowlist excludes this model. Both are fixed on the tokens page.

Nodes time out occasionally

Platform defaults are often 60 seconds. Raise the timeout on long-output nodes or switch to streaming. The model detail page shows 24-hour latency percentiles.

Costs higher than expected

Filter the console logs by model and key. Workflows often call a node repeatedly inside a loop. A dedicated key with a spend cap keeps a workflow in check.

Cost example

At the current price of qwen3.8-flash, one typical request with 1,500 input + 300 output tokens, 5,000 times a day:

You pay only for tokens actually used. No platform fee, routing fee or monthly minimum. Real usage is listed request by request in the console logs.