Read the allowlist literally
Alibaba's Token Plan Individual documentation tells users to match model IDs character for character. A different date, version, suffix or sub-variant is not supported unless it is separately listed. Brand familiarity is not enough: access to one Qwen, Wan, GLM, DeepSeek or HappyHorse model does not imply access to every model from that provider.
The roster below reflects the official Individual overview checked for this draft on 23 July 2026. Preview models, third-party models and promotional Credit coefficients can change. Production copy should retain a visible verification date and direct readers to the official allowlist before they change a working configuration or purchase a plan.
Qwen text, reasoning and visual understanding
Visual understanding means a supported model can reason over visual input; it is different from generating a new image. Qwen3.8 Max is explicitly a preview model. Alibaba says it may be taken offline or replaced with a production version after preview, so integrations should use its exact ID and avoid assuming a permanent alias.
Alibaba currently advertises limited-time reduced Credit consumption for qwen3.8-max-preview, including an additional reduction for calls between 22:00 and 08:00 UTC+8. The offer can be adjusted, and it should not be described as a permanent plan capacity.
- qwen3.8-max-preview - reasoning, visual understanding and text generation.
- qwen3.7-max - reasoning and text generation.
- qwen3.7-plus - reasoning, visual understanding and text generation.
- qwen3.6-flash - reasoning, visual understanding and text generation.
Third-party reasoning and text models
These IDs demonstrate that Individual is multi-provider, but they remain exact entries rather than access to the full GLM or DeepSeek catalogs. A tool configuration should copy the lowercase model ID from the current table. Inferring support for a Flash version, dated snapshot or later major release can return a model-not-supported error.
- glm-5.2 from Zhipu AI - reasoning and text generation.
- deepseek-v4-pro from DeepSeek - reasoning and text generation.
Wan image-generation models
Image-generation models are not selected in a coding tool in the same way as a chat model. Alibaba's multimodal guide says coding tools must call them through an extension mechanism such as a Skill, Slash Command or Agent. The Token Plan multimedia endpoint and plan key are used by that extension while the human remains in an interactive tool session.
That integration detail prevents a common misunderstanding: a model can be included in Individual without appearing as a normal conversational model in a tool's model picker. The tool must support an extension path capable of making the documented media request.
- wan2.7-image - image generation.
- wan2.7-image-pro - image generation.
HappyHorse 1.1 video workflows
Alibaba documents video generation as an asynchronous workflow. An extension submits a task, stores the returned task ID, polls its status and downloads the file after success. The current example supports five-second default duration, 720P default resolution, optional 1080P and workflow-dependent aspect-ratio inputs.
The supported roster does not publish one fixed number of videos per tier. Credit consumption depends on the model and workload, so an exact allowance should come from current subscription usage details and the chosen parameters, not from dividing a headline Credit balance by an assumed cost.
- happyhorse-1.1-t2v - text-to-video from a written description.
- happyhorse-1.1-i2v - image-to-video from a first-frame image plus text.
- happyhorse-1.1-r2v - reference-to-video from reference images plus text.
Harness tools expand a conversation
Harness-tool calls also consume Credits according to their deduction coefficient. Alibaba's quick start currently says native Harness tool calling through the Responses API is supported by the Qwen3.7 and Qwen3.8 series. The presence of a tool in the plan does not mean every included model can call it natively.
- web_search for web search.
- t2i_search for text-to-image search.
- i2i_search for reverse image search.
- web_extractor for web scraping.
- code_interpreter for code execution and analysis.