# Qwen: compare connected agent work, frontend building and local execution Inspect the task and output at each stage instead of treating the cloud trial and local model as the same route. [HTML](https://lukejbyrne.com/library/guides/rrqlt8mm1oc/) Content language: en Author: Luke Byrne Video published: 2026-05-07 Guide published: 2026-11-02 [Watch the video](https://www.youtube.com/watch?v=Rrqlt8Mm1oc) Video language: English ## Connect the cloud route for a bounded task The May description links an Alibaba Cloud trial: https://int.alibabacloud.com/se/1023344164/ advertised as 70M free tokens. That allowance belongs to the original offer; check current eligibility and terms at the retained destination. The recorded sequence gets a cloud API key and connects Qwen inside a coding IDE. It then explores four jobs rather than establishing a universal benchmark ranking: an agentic workflow, a frontend page, a multi-step app and a local open-weight model. 1. Follow the trial section at 1:03 and IDE connection at 2:21. 2. Choose a small task and inspect the selected model route before starting. 3. Keep cloud usage distinct from later local execution when comparing results and cost. ## Build the packaging agent and landing page The first test creates a YouTube packaging agent at 3:31, adds transcript API tools at 5:50 and tests the improved agent at 6:44. Thumbnails with Qwen Image Max follow at 8:03. This progression makes the inputs and connected tools part of the agent’s result. 1. Check the source video and transcript material used by the packaging task. 2. Review the generated wording and thumbnail against that material. 3. Follow the frontend build at 8:58 and page comparison at 10:30. 4. Compare the design-skill revision at 11:49 with the original page and its intended visitor task. ## Keep the app test separate from the local-model test TweetCaster starts at 13:32, followed by testing at 15:32 and interface/output improvements at 16:13. Inspect the application’s behaviour before switching the model, so you can distinguish an app change from a model-route change. The offline section starts at 17:49, installs Qwen with Ollama at 18:13 and switches the app at 20:00. An open-weight local model is a different setup from the earlier cloud API. Use the offline test section at 20:54 to assess the reported result and inspect any remaining network dependencies when adapting the app. ## The takeaway Compare each result with its own input and model route; cloud tokens, connected tools and a local Ollama model serve different parts of the workflow. ## Workflow resources - [int.alibabacloud.com](https://int.alibabacloud.com/se/1023344164/) ## Chapters - [0:00 Alibaba Qwen is blowing up](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=0s) - [1:03 How to get free Alibaba Cloud AI tokens](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=63s) - [2:21 Connecting Qwen to a coding IDE](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=141s) - [3:31 Test 1: Building a YouTube packaging agent](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=211s) - [5:50 Adding transcript API tools](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=350s) - [6:44 Testing the improved agent](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=404s) - [8:03 Adding thumbnails with Qwen Image Max](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=483s) - [8:58 Test 2: Building a frontend landing page](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=538s) - [10:30 Comparing the generated landing page](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=630s) - [11:49 Improving the UI with a design skill](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=709s) - [13:32 Test 3: Building TweetCaster](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=812s) - [15:32 Testing the generated app](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=932s) - [16:13 Improving the TweetCaster UI and output](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=973s) - [17:49 Test 4: Running Qwen offline](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=1069s) - [18:13 Installing Qwen locally with Ollama](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=1093s) - [20:00 Switching the app to a local model](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=1200s) - [20:54 Offline test results](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=1254s) - [21:20 Final thoughts](https://www.youtube.com/watch?v=Rrqlt8Mm1oc&t=1280s) ## Sources and limitations Written companion prepared 2 November 2026 from this exact published video description and its chapter sequence. The full recording was not independently watched or transcribed for this article, and the products and reported results were not retested. Checklists are editorial adaptations of the documented workflow. Model availability, interfaces, prices and offers may have changed since publication; use the retained original resources for current requirements. [Original video](https://www.youtube.com/watch?v=Rrqlt8Mm1oc)