Can MiniMax H3 Work with DSH?
Learn how MiniMax H3 differs from M-series coding models, which API routes DSH uses, and what to verify before integrating a video asset.
On this page
MiniMax H3 demos can make it tempting to paste its name into DSH's model list. H3 targets multimodal generation and video creation. For a coding agent, start with MiniMax's M-series text and tool-calling interfaces. The two can contribute to one workflow, but changing a model ID does not make their interfaces interchangeable.

How MiniMax H3 differs from the M series
MiniMax's H3 announcement, dated July 31, 2026, describes multimodal references and generation involving images, video and audio. Those capabilities concern creative assets; they do not establish that H3 can act as a coding model through a chat endpoint. Source: MiniMax H3 announcement.
The Anthropic-compatible text API lists MiniMax-M3, M2-series models and MiniMax-M3.1-Flash-Preview. It supports text and tool-related messages, but does not list H3 as a chat model. Source: MiniMax Anthropic SDK guide.
Choose by the deliverable, rather than asking which family wins overall:
| Task | Starting point | Acceptance check |
|---|---|---|
| Explain code, produce patches or use development tools | M-series text API and DSH model configuration | File changes, tests and tool-call records |
| Create a video from reference material | H3 creation interface and its generation documentation | Video file, shots, subject consistency and audio |
| Produce assets and add them to a product page | Separate asset generation from code integration | Approve the asset, then check its path and rendering |
A DSH connection needs a protocol, not just a model name
MiniMax documents these text routes for its international platform. A base URL is a service prefix; parameters for different protocols should not be combined casually.
| Route | Base URL | Matching DSH custom-provider protocol |
|---|---|---|
| OpenAI-compatible Chat Completions | https://api.minimax.io/v1 | openai-completions |
| Anthropic-compatible Messages | https://api.minimax.io/anthropic | anthropic-messages |
The endpoints come from the OpenAI SDK guide and Anthropic SDK guide. DSH's protocol names come from its model configuration guide at source snapshot c291e79. Do not apply the international endpoints to credentials from another region without checking that platform's documentation.
If your installation already offers a suitable MiniMax provider, start there. If its catalog lacks the model you need, check the DSH custom-provider guide. Keep a working text provider intact rather than replacing its address with an unverified H3 route.
Verify a minimal text request first, then a read-only tool task. Only then try edits in a project copy. A successful chat response does not prove tool calling or acceptable code changes.
A saved key does not establish model access
MiniMax distinguishes pay-as-you-go API keys from subscription keys. Its documentation currently uses names including M Plan and Token Plan, while retaining references to the former Coding Plan. Check the resource rules in the account and plan you actually use. Source: API prerequisites.
The Token Plan guide says subscription keys are not interchangeable with pay-as-you-go keys. It also excludes certain special models, including H3, from the plan's current coverage. A coding or multimodal subscription is therefore not evidence that every MiniMax model is available. Source: Token Plan overview.
The text API guide adds another access condition: MiniMax-M3.1-Flash-Preview is currently available only through M Plan and MiniMax Code. Check eligibility before copying that model from an example into a pay-as-you-go integration. Source: OpenAI SDK guide.
For troubleshooting, record the platform and product that issued the key, the exact model ID, and the resources available to the account. Never include the key itself in screenshots. Guessing a newer model name is not a substitute for those checks.
Divide video production and code integration into separate steps
For a product-page video, a coding agent can help turn approved product information into a shot brief. Use a confirmed H3 entry point to generate and review the asset. Then ask DSH to modify the page component, reference the saved file and run the relevant checks.
Separate acceptance checks make failures easier to locate. An inaccurate product in the video is an asset problem. An approved video referenced at the wrong path is an integration problem. A single instruction to “make and publish a video” can hide both.
Automation needs a tool that calls the video service, handles job status and saves the output. DSH can connect to external MCP tools. Before choosing a service, check its published tool definitions, authentication and H3 support. If those details are missing, generate and save the video manually, then let DSH integrate it into the page. Source: DSH MCP client reference at c291e79.
DSH's installed catalog and provider APIs update independently. This guide uses the cited source snapshot and provider documentation accessed on October 2, 2026. We have not completed an end-to-end test with H3 or MiniMax API credentials.
FAQ
Can I enter H3 as a DSH text-model ID?
The reviewed MiniMax text API model list does not include H3. Choose a listed model and check that it matches the protocol and output type.
Why does visual input in M3 not establish video generation?
Understanding an input image and returning a generated video are different capabilities. Check the output format and supported tools separately instead of treating “multimodal” as a promise of every media capability.
Do I need a subscription to use MiniMax?
MiniMax also offers pay-as-you-go API access. Choose the model and product before choosing a billing route. Conditions for one preview or special model are not universal rules for all models.
Do I need MCP for a single product video?
You can start without it. Generate and review the video, then pass the saved asset to DSH for page integration. If you need to repeat that workflow, check whether a suitable MCP tool is available.
Next step
Before using MiniMax H3, decide whether the deliverable is a video asset or a code change. Start coding work with the MiniMax provider guide. For video work, confirm H3 access and approve the result before integrating it into a DSH workflow.
Model and workflow guides
Connect Qwen 3.8 27B to DSH
Connect local Qwen 3.8 27B to DSH: check the server, model ID, protocol and credentials, then test text, tools and image input separately.
Read →Model setup and costsCodex Costs: API Billing vs Subscription Usage
Understand Codex API billing, subscription limits and credits, with GPT-5.6 Sol vs GPT-5.5 rates and a clearly scoped token-cost example.
Read →Model setup and costsConnect the Grok API to DSH
Connect the Grok API to DSH: distinguish Grok from Groq, match credentials and protocol, and check independent requests, text and tools.
Read →Model setup and costsChoose a Gemini Model by Task and Cost
Choose Gemini by task, inputs, interface and API cost. Check promotion dates and model status, then verify what your DSH adapter supports.
Read →Model setup and costsClaude Sonnet vs Opus: Which Should You Use?
Compare Claude Sonnet and Opus 5.5 by API rates, retries and review effort, then use the same coding task to decide whether an upgrade is worthwhile.
Read →API credentials and troubleshootingKimi API Key Errors: Region, Model and Permissions
Fix Kimi API key setup errors by matching region and model ID, reading HTTP responses, and checking DSH or OpenChamber's actual configuration.
Read →Image editing and toolsAI Image Editing: Prompts, Masks and Review
Learn to define image edits, preserve important details, check masks and output files, and verify an editing tool before organizing it with DSH.
Read →