Bring a document, a code excerpt or a complete task brief into a conversation with Kimi K3. Describe the result you need, provide the relevant source text and specify the structure of the answer. The existing Vidgo task uses Chat Completions and provides system instructions, conversation messages, output-token limits and sampling controls. Review the returned text and token usage, then continue with clarifications or test feedback when refining the result.
Kimi is developed by Moonshot AI. Vidgo provides access to the models. Model names and trademarks belong to their respective owners.Use the existing Kimi K3 task to enter your goal and source material. Start a document-analysis conversation, request a structured answer or provide code with the behavior you want to inspect.
Write the instruction as a user message and include the document excerpt or code needed for the task. Use a system instruction for persistent role or formatting guidance, and keep relevant earlier user and assistant messages when continuing the conversation. Set an output-token limit that fits the requested answer. The task uses model: kimi-k3 at /v1/chat/completions.
| Model / task | Input and workflow |
|---|---|
| Kimi K3 | Moonshot AI Kimi K3 supports a 1,048,576-token context for repository analysis, long-document workflows, coding, and sustained agent planning. |
Kimi K3 supports sustained coding, task planning and professional knowledge work. Its published charts cover coding benchmarks, general and visual agent evaluations, and internal knowledge-work tests. The results correspond to the stated evaluation environments.
The coding chart covers repository changes, terminal tasks and longer development workflows across the listed benchmarks. Provide the project context, relevant code, desired behavior and test requirements so Kimi K3 can help plan a change, write an implementation and explain how to verify the result.
The agent chart reports performance on tasks that combine planning, information handling and, in some tests, visual understanding. For your workflow, include the material already collected and describe the required steps and output. Connect the model to your own tools when the task requires external actions or verification.
The knowledge-work chart covers internal tests for research tasks, presentation content and financial analysis. Supply the relevant reports, business context and intended audience, then request an organized analysis, a reasoned conclusion or a presentation outline that can be developed into your final deliverable.
Prepare the information the model needs and explain how the answer will be used. Ask for a summary, structured extraction, code explanation or a complete draft, then review the result against your sources and project requirements. Continue the conversation with concrete corrections and test feedback.
Provide the relevant files or excerpts, explain the repository task and ask for the responsibilities and relationships to inspect. Review the explanation against the code before using it in a technical handoff or implementation discussion.
Prepare the relevant document sections and request a summary with named topics, facts and follow-up questions. Check each point against the source when creating a working note for further analysis.
Provide a source note and define the required fields, units and output structure. Validate the extracted values before using the result in a catalog draft, structured content workflow or application prototype.
Supply the source facts, intended reader and required sections, then request a complete explanation or reply. Review the wording and retained details before using the draft in a document or customer workflow.
Provide the code, expected behavior and observed test output. Ask for the next proposed change and verify it with your tests, then return the result to the conversation when another iteration is needed.
Describe the goal, available material and completion criteria, then keep relevant turns when updating the plan. Add new findings and explicit constraints so the next answer can build on the project’s current evidence.
The table reads Kimi K3’s current starting input-token rate per one million tokens. Its task page lists input, output, cache-read and cache-write rates. The conversation history sent with a request contributes to input usage, while the generated answer contributes to output usage. Review the relevant rates together with the returned usage information when planning a document-analysis conversation or several coding iterations.
| Model | Provider | Workflow | Input | API formats | Starting input rate |
|---|---|---|---|---|---|
| Kimi K3 | Moonshot AI | Chat | Text instructions and conversation context | Chat Completions | $2.28 / 1M tokens |
Test a complete text task online, then integrate the selected model and API format.
Open the selected model, provide the task and source text, and choose its available output settings. Submit the conversation, read the complete answer and inspect available token usage. Add clarifications or test feedback when preparing a follow-up.
Use the existing model conversation, current token prices and task documentation to connect a reviewed answer with your application workflow.
Open Kimi K3 and prepare a text task around the document, code or brief you already have. Choose Chat Completions, describe the required answer and keep the relevant source material in the conversation.
Open the selected model, provide the task and source text, and choose its available output settings. Submit the conversation, read the complete answer and inspect available token usage. Add clarifications or test feedback when preparing a follow-up.
The table reads Kimi K3’s current starting input-token rate per one million tokens. Its task page lists input, output, cache-read and cache-write rates. The conversation history sent with a request contributes to input usage, while the generated answer contributes to output usage. Review the relevant rates together with the returned usage information when planning a document-analysis conversation or several coding iterations.
Send a model identifier and messages to /v1/chat/completions. A user message carries the task and source text; a system message can carry persistent guidance. The online task supports streaming text, so your application can accumulate response chunks and inspect the final usage information.
Use your Vidgo API credentials to authenticate the request to the selected endpoint. Send the exact model identifier, the conversation fields required by that protocol and the supported output settings. Consume the returned response using the same format you selected.
Accumulate streaming text when the selected workflow streams, or read the complete response body when it returns a full reply. Inspect the final answer and returned usage fields, then keep the relevant messages for the next turn. Return to the model conversation when continuing the online task.
Bring a document, a code excerpt or a complete task brief into a conversation with Kimi K3. Describe the result you need, provide the relevant source text and specify the structure of the answer. The existing Vidgo task uses Chat Completions and provides system instructions, conversation messages, output-token limits and sampling controls. Review the returned text and token usage, then continue with clarifications or test feedback when refining the result.
Read language-model introductions, API guides and platform comparisons to understand model selection, conversation inputs, response formats and token costs.
