What does Token Forge Cloud do?

Token Forge Cloud provides managed APIs for Qwen, DeepSeek, GLM, and Seedance, with additional video and speech models for teams building production AI products. Proven workloads can also move into a Neo Cloud marketplace product path.

Who is Token Forge Cloud built for?

It is built for product and enterprise teams that need reliable language, reasoning, coding, agent, and video model access without operating every provider stack themselves.

Can teams start with APIs before buying through Neo Cloud?

Yes. Teams can begin with managed API access for Qwen, DeepSeek, GLM, Seedance, and additional media models, then move predictable production traffic into a Neo Cloud marketplace product path.

Which AI workflows does Token Forge Cloud support?

Token Forge Cloud supports assistants, reasoning, coding, agents, text-to-video, image-to-video, asynchronous jobs, model routing, and speech generation.

How does Token Forge Cloud handle security and data control?

Token Forge Cloud supports role policies for who can call which models and workflows, controlled access for prompts and application data, and audit-ready request, generation, and routing telemetry.

MiniMax API: Access, Model Selection and Integration

MiniMax uses different models and API routes for text, speech, and video. For chat, the official text guide covers models such as MiniMax-M3.1-Flash; text-to-speech uses a separate endpoint and speech model; video generation creates an asynchronous task. Don’t assume one model, credential, or request format works across all three. The examples and links below are tied to the specific MiniMax documentation they describe.

Choose a model by task

Task	Model options identified in MiniMax documentation	What to check
Text and chat	`MiniMax-M3.1-Flash`, `MiniMax-M3`	The text guide describes M3.1-Flash as a 1-million-token-context model with tunable thinking effort. Use the model and request format supported by your chosen integration. Text-generation guide
Text-to-speech	`speech-2.8-hd`	MiniMax’s synchronous TTS reference uses this model. Check the official model list for other listed speech models. Model list · TTS API reference
Video generation	`MiniMax-H3`; the model list also identifies Hailuo video models	H3 task creation is documented separately from chat. Check the video guide for the generation mode and constraints that apply to your task. Video-generation guide

What about MiniMax 2.5?

MiniMax’s model-list notice gives October 19, 2026 as a cutoff for new-user access and November 18, 2026 as the retirement date for existing pay-as-you-go users. Confirm that the notice applies to the exact M2.5 model ID or variant you use and to your account plan before relying on it or planning a migration; its applicability beyond the stated pay-as-you-go users is not clear here. MiniMax model list and notice

Access: credentials and endpoints

MiniMax distinguishes a pay-as-you-go API key, created under API Keys, from an M Plan subscription key, found under Billing → M Plan. These are different credential types; don’t assume one can replace the other for every model or endpoint. Confirm the eligible credential and account requirements in the relevant API guide before making a request. Quickstart prerequisites · API overview

Task	Documented route	Applicability
Text chat	MiniMax documents Anthropic-compatible and OpenAI-compatible chat routes. The text guide gives the applicable base URLs and request examples.	Follow the format and credential instructions for the SDK or API route you choose; don’t combine settings from different formats. Text-generation guide
Synchronous text-to-speech	`POST https://api.minimax.io/v1/t2a_v2`	This is the documented HTTP TTS endpoint. Use the TTS reference for its required body and response fields. TTS API reference
H3 video generation	`POST https://api.minimax.io/v2/video_generation`	The creation reference describes an asynchronous task and a returned task ID. The eligible credential and any account prerequisite for this H3 route are not established here; confirm them in the current video guide before choosing a key. Create-video reference · Video-generation guide

The URLs above are the global hosts shown in the linked MiniMax documentation. The materials available here do not establish that the same host, credential, or plan requirements apply to every region and modality.

First integration step

There is no single verified request body that covers chat, speech, and video. Start with the official reference for the selected task: the text-generation guide, synchronous TTS reference, or H3 task-creation reference. Each provides a modality-specific example; use its required fields and credential instructions rather than copying a shared template.

If you want a named client setup rather than a raw API request, MiniMax documents this Codex setup:

Obtain an M Plan API key from Plan Details. The guide recommends Codex CLI 0.146.0 or newer.
Close Codex and any editor that has ~/.codex/config.toml open.
Run the setup wizard and select Codex:
```
npx -y mmx-cli@latest agent setup
```
Restart Codex after setup. The wizard configures Codex configuration and model-catalog files. Use /model to inspect the available model and reasoning choices; the guide describes an optional catalog configuration for showing MiniMax-M3.1-Flash and its supported reasoning levels. Seeing a model in that list is a configuration check, not proof that an API request succeeded.

This path is specifically for Codex and MiniMax-M3.1-Flash; its manual configuration uses an M Plan API key, https://api.minimax.io/v1, and the Responses wire format. It is not a general-purpose request example for speech or video. MiniMax Codex guide

Video task completion

The H3 creation step returns a task_id; it does not return the finished video. This article covers task creation only. Use MiniMax’s video-generation guide for the status and result workflow; the exact query parameters and result fields are not reproduced here.

Which route should you use?

Chat or coding: Start with the text-generation guide and select a listed chat model such as MiniMax-M3.1-Flash. Its guide describes a 1M-token context window and tunable thinking effort; follow the guide’s exact API format and request example.
Speech from text: Use the TTS endpoint and a speech model such as speech-2.8-hd, not the chat endpoint. Choose the synchronous HTTP reference for that request; consult MiniMax’s speech documentation for other modes.
Video creation: Use the video-generation API and follow the task-based workflow. The eligible credential and account requirements for H3 remain unconfirmed here, so verify them in the current official guide before choosing a key.
An existing M2.5 integration: Check the model-list notice against your account type and exact variant. It gives November 18, 2026 as the retirement date for existing pay-as-you-go users; confirm whether that notice covers your precise model ID and plan before making a migration decision.

Common integration pitfalls

Don’t mix chat formats. Use the base URL, headers, and body from the same MiniMax compatibility guide.
Don’t use the chat endpoint for TTS or video. TTS has its own /v1/t2a_v2 endpoint; H3 task creation uses /v2/video_generation.
Don’t assume credentials are interchangeable. MiniMax lists pay-as-you-go and M Plan credentials separately. Check the selected model’s guide for eligibility.
Don’t treat video task creation as completed generation. Keep the task ID and follow MiniMax’s status and result instructions in the video guide.
Keep API keys out of source code. MiniMax’s quickstart prerequisites explain where to obtain the relevant credential and recommend using an environment variable. Quickstart prerequisites

Official documentation

Sources

Model List - MiniMax API Docs

API Overview - MiniMax API Docs

Prerequisites - MiniMax API Docs

Model Invocation - MiniMax API Docs

Synchronous Text to Speech - MiniMax API Docs

Video Generation - MiniMax API Docs

Create Video Generation Task - MiniMax API Docs

Codex - MiniMax API Docs