Choose the workflow you actually need
Work in your editor
Cline →Use the verified connection walkthrough, then test a small file edit separately.
Code from your terminal
OpenCode →Select the built-in IteraCompute provider and try the documented free Qwen repair workflow.
Run coding tasks in VS Code
Roo Code →Configure an OpenAI Compatible provider, then check file edits and test commands on a small task.
Call from an application
OpenAI SDK →Use Python or JavaScript on the server. Check non-empty text before adding streaming or tools.
Build a TypeScript AI feature
Vercel AI SDK →Start with the tested server-side CLI example. Hosting and browser UI need separate verification.
Understand the Auto allowance included with your plan
The free promotion ends at 00:00 Pacific Time on November 24, 2026. Users with positive net recharge completed before this cutoff retain their existing free tier. Afterward, shared free Auto is limited to these users and users with a currently effective plan. Later recharges do not grant or upgrade free benefits. Plan Auto tokens and monthly shared free benefits are separate allowances.
Subscription plans include a separate Auto token allowance alongside paid-model usage. The Auto allowance varies by plan; check the plans page and checkout for current amounts and periods.
The Auto plan allowance, shared paid-model allowance and account free benefits are counted separately. Auto requests use the plan allowance first. Requests that fall back to free access use shared capacity, which may involve queues or slower responses at peak times.
Run a task with a clear finish line
For a coding evaluation, use an empty folder and a disposable Git branch. Ask your tool to make two files and run the tests. Review any requested file and command actions.
Create slugify.mjs exporting slugify(value), and slugify.test.mjs using node:test.
Convert to lowercase, trim whitespace, replace runs of non-ASCII-alphanumeric characters with one hyphen, and remove leading/trailing hyphens.
Test these exact cases:
" Hello, World! " -> "hello-world"
"API Key" -> "api-key"
"---" -> ""
"" -> ""
Run node --test slugify.test.mjs and report the result. Do not install dependencies.- Connection passes only when the tool returns visible non-empty text.
- The coding task passes only when both files exist, the diff matches the request and all four tests pass with exit code 0.
- Record model ID, tool version, date, input/output tokens, elapsed time and failures. A single run is an example, not a quality benchmark.
For document Q&A, use a short document with five answers you already know. Ask for source quotations and mark each answer correct, unsupported or missing. For a chat feature, test five representative user requests and one request it should decline or hand off.
Diagnose the bottleneck before changing models
Empty reply or connection error
Check credentials, endpoint and output budget using the tutorial. A larger paid model does not fix a wrong configuration.
429 or a full context window
Check monthly usage, any queued or running request, and input plus requested output. Wait for the earlier request to finish or shorten the conversation; compare tiers for monthly allowance and context.
Connection works, task still fails
Repeat the same task on a paid candidate, with the same inputs and acceptance criteria. Compare the actual result and cost.
Predictable production capacity
Discuss reserved capacity or dedicated deployment. A Qwen recharge does not provide a latency SLA.
Compare the cost of your actual workload
Check current prices and specifications in the model catalog. Subscription plans provide more usage; see the plans page for allowances, eligible models and usage rules. For predictable production capacity, explore reserved capacity.