MiniMax-Text-01: Complete Specifications, Pricing, API Access & Use Cases (2026)
What Is MiniMax-Text-01?
MiniMax-Text-01 is MiniMax’s open-weight long-context language model, released on January 15, 2025, featuring a 1M-token training context, up to 4M-token inference context, text input, text output, and launch API pricing of $0.20 per 1M input tokens and $1.10 per 1M output tokens as of June 2026.
MiniMax released MiniMax-Text-01 as part of the MiniMax-01 series, together with MiniMax-VL-01, a related vision-language model built on top of the text model. The text model uses a hybrid architecture that combines Lightning Attention, Softmax Attention, and Mixture-of-Experts routing. Official model materials list 456B total parameters, 45.9B activated parameters per token, 80 layers, 32 experts, and top-2 expert routing as of June 2026.
Users usually search for MiniMax-Text-01 because it is associated with unusually long context handling, open-weight access, and comparison against long-context or high-capability models such as Gemini 1.5 Pro and Claude 3.5 Sonnet. It should be treated as a text-focused long-context LLM, not as the multimodal model in the MiniMax-01 family.
What Are MiniMax-Text-01’s Key Specifications and Pricing?
| Field | Verified Value |
|---|---|
| Provider | MiniMax, as of June 2026 |
| Model Family | MiniMax-01 series, as of June 2026 |
| Model Type | Open-weight long-context language model / hybrid-attention MoE LLM, as of June 2026 |
| Release Date | January 15, 2025 announcement; model license lists January 15, 2025 release date, as of June 2026 |
| Context Window | 1M tokens during training; up to 4M tokens during inference, as of June 2026 |
| Input Pricing | Launch announcement: $0.20 per 1M input tokens; current MiniMax pay-as-you-go page does not list MiniMax-Text-01 pricing, as of June 2026 |
| Cached Input Pricing | Not confirmed from official MiniMax-Text-01 sources as of June 2026 |
| Output Pricing | Launch announcement: $1.10 per 1M output tokens; current MiniMax pay-as-you-go page does not list MiniMax-Text-01 pricing, as of June 2026 |
| Pricing Unit | Per 1M tokens in the launch announcement, as of June 2026 |
| Modality Support | Text input and text output for MiniMax-Text-01; MiniMax-VL-01 is the vision-language model in the same series, as of June 2026 |
| Supported Input Types | Text, as of June 2026 |
| Supported Output Types | Text, as of June 2026 |
| API Access | MiniMax announced Open Platform API access at launch, but current MiniMax text-generation docs list MiniMax-M3, MiniMax-M2.7, MiniMax-M2.5, MiniMax-M2.1, and MiniMax-M2 as selectable model IDs rather than MiniMax-Text-01, as of June 2026 |
| Model ID | Open-weight repository/model ID: MiniMaxAI/MiniMax-Text-01; current selectable MiniMax API model ID is not confirmed as of June 2026 |
| Availability | Open weights via official MiniMax GitHub and Hugging Face materials; current hosted API availability for this exact model is unclear, as of June 2026 |
| Knowledge Cutoff | Not confirmed from official sources as of June 2026 |
| Rate Limits | Not confirmed from official MiniMax-Text-01 sources as of June 2026 |
| Fine-Tuning Support | Not confirmed from official sources as of June 2026 |
| Streaming Support | Current MiniMax text-generation endpoint supports streaming, but the current selectable model list does not include MiniMax-Text-01 as of June 2026 |
| Batch API Support | Not confirmed from official sources as of June 2026 |
| Tool / Function Calling | Current docs describe tool fields for MiniMax text-generation APIs, but MiniMax-Text-01-specific tool-calling support is not confirmed as of June 2026 |
| Structured Output / JSON Mode | MiniMax docs state response_format is only supported by MiniMax-Text-01; exact current selectable API availability remains unclear, as of June 2026 |
| License / Usage Restrictions | MiniMax Model License applies to model materials; code license is MIT. The model license includes redistribution, attribution, legal compliance, prohibited-use, warranty, and liability terms, as of June 2026 |
What Can MiniMax-Text-01 Do That Makes It Useful in Production?
MiniMax-Text-01’s main production value is long-context text processing. Official materials state that the model was trained with a 1M-token context and can extrapolate to up to 4M tokens during inference, making it relevant for long documents, code repositories, research corpora, transcript archives, and agent-memory experiments. The important limitation is that long context does not remove the need for retrieval quality, prompt filtering, output validation, or human review.
The model is also useful for infrastructure teams evaluating open-weight long-context systems. MiniMax provides model materials through its official GitHub and Hugging Face presence, which can support controlled deployment research, inference optimization, and internal evaluation. This may matter for teams that want more deployment control than a hosted-only model, but it also increases responsibility for hardware, security, monitoring, and license compliance.
MiniMax-Text-01 may fit structured extraction workflows because MiniMax’s current text-generation documentation states that response_format is only supported by MiniMax-Text-01. This makes the model relevant for JSON-style extraction and schema-constrained outputs, but production users should verify whether MiniMax-Text-01 is currently selectable in their API account before building against that capability.
The model is also useful as a benchmark target for long-context evaluation. MiniMax’s model card compares MiniMax-Text-01 against models including GPT-4o, Claude 3.5 Sonnet, Gemini 1.5 Pro, Gemini 2.0 Flash, Qwen2.5-72B Instruct, DeepSeek-V3, and Llama 3.1 405B Instruct. These benchmarks can guide initial evaluation, but they should not replace task-specific tests on private data, especially for legal, medical, financial, compliance, or safety-sensitive workflows.
What Are MiniMax-Text-01’s Supported Modalities?
| Modality | Supported? | Notes |
|---|---|---|
| Text input | Yes | MiniMax-Text-01 is described as a language model |
| Text output | Yes | The model is designed for text generation and long-context language tasks |
| Image input | No for MiniMax-Text-01 | MiniMax-VL-01 is the related vision-language model built from MiniMax-Text-01 |
| Audio input | Not confirmed | No MiniMax-Text-01-specific audio input support was found in official sources |
| Video input | Not confirmed | No MiniMax-Text-01-specific video input support was found in official sources |
| Tool / function calling | Not confirmed as model-specific | Current text-generation docs include tool fields, but not as a confirmed MiniMax-Text-01-specific capability |
| JSON / structured output | Partially confirmed | MiniMax docs state response_format is only supported by MiniMax-Text-01 |
Where Does MiniMax-Text-01 Fall Short?
MiniMax-Text-01’s most important limitation is current access clarity. The January 2025 launch announcement verified Open Platform API access and launch pricing, but the current MiniMax pay-as-you-go page lists MiniMax-M3 and M-series models rather than MiniMax-Text-01. Current text-generation API docs also list newer selectable model IDs and mention MiniMax-Text-01 only in some parameter notes.
MiniMax-Text-01 is not the multimodal model in the MiniMax-01 series. Image understanding belongs to MiniMax-VL-01, which MiniMax describes as a vision-language model built on top of MiniMax-Text-01. Teams that need image, audio, or video understanding should not assume those modalities are supported by MiniMax-Text-01.
Long context is not the same as guaranteed long-context reasoning. This is a general AI limitation and is not model-specific unless stated by the provider. Even with very large context windows, long prompts can include irrelevant material, stale facts, contradictory evidence, or hidden prompt-injection risks. Production systems should still use retrieval ranking, content filtering, evaluation sets, logging, and human escalation.
Open-weight deployment can be operationally demanding. The official model size and architecture imply substantial inference requirements, especially for long-context workloads. Teams should evaluate GPU memory, inference framework support, quantization trade-offs, latency, and data-protection controls before production deployment.
License compliance also matters. The MiniMax Model License grants a limited license for model materials but includes redistribution duties, attribution requirements, prohibited-use obligations, warranty disclaimers, and other restrictions. The separate code license is MIT, but that does not make the model weights unrestricted.
What Is MiniMax-Text-01 Best Used For?
| Use Case | Why MiniMax-Text-01 May Fit | Important Limitation |
|---|---|---|
| Long-document analysis | The model is designed around very long text context, with official sources stating up to 4M inference tokens | Large prompts still need relevance filtering and verification |
| Research-paper and technical-corpus review | The long context window can help keep more source material in a single request | Generated summaries should be checked against original sources |
| Codebase and documentation review | Long context can support larger multi-file or documentation-heavy tasks | Code correctness still requires tests, static analysis, and human review |
| Structured text extraction | MiniMax docs reference response_format support for MiniMax-Text-01 |
Current selectable hosted API availability should be verified before production use |
| Open-weight long-context evaluation | Official materials expose the model for research and deployment evaluation | Hardware, license, and security controls can be significant |
| Comparative LLM benchmarking | Official MiniMax materials compare against models such as Qwen2.5-72B Instruct and Llama 3.1 405B Instruct | Public benchmark results should be supplemented with task-specific testing |
How Does MiniMax-Text-01 Compare to Gemini 1.5 Pro and Claude 3.5 Sonnet?
| Comparison Area | MiniMax-Text-01 | Gemini 1.5 Pro | Claude 3.5 Sonnet | Scenario Fit |
|---|---|---|---|---|
| Provider | MiniMax, as of June 2026 | Google, as of June 2026 | Anthropic, as of June 2026 | Choose based on deployment path, data policy, ecosystem, and current availability |
| Model Type | Open-weight long-context text LLM | Hosted long-context Gemini model | Hosted Claude model | MiniMax may fit open-weight evaluation; Gemini and Claude may fit managed API workflows |
| Context | 1M training context and up to 4M inference context in official MiniMax materials, as of June 2026 | Public Gemini 1.5 Pro sources have described very long context support; verify the current Gemini version and tier before use | Anthropic announced Claude 3.5 Sonnet with a 200K context window | MiniMax is especially relevant when ultra-long text context is the core evaluation criterion |
| Pricing | Launch price was $0.20 input and $1.10 output per 1M tokens; current MiniMax pay-as-you-go page does not list MiniMax-Text-01 | Gemini pricing varies by model version and platform; verify the current provider page | Claude pricing varies by version and platform; verify the current provider page | Pricing comparisons should use current provider or gateway pricing at purchase time |
| Modalities | Text input and text output for MiniMax-Text-01 | Gemini models are commonly used for multimodal workflows depending on version | Claude 3.5 Sonnet supports text and vision workflows depending on access route | Use MiniMax-Text-01 for text-first long context; use multimodal models for image tasks |
| Deployment | Open-weight model materials are available; hosted API availability for this exact model is unclear in current MiniMax docs | Hosted Google ecosystem | Hosted Anthropic ecosystem and partner access | Deployment preference may be more important than benchmark differences |
How Do I Access MiniMax-Text-01?
MiniMax-Text-01 can be accessed through official open-weight materials published by MiniMax through its GitHub and Hugging Face model resources. The verified open-weight model identifier is MiniMaxAI/MiniMax-Text-01 as of June 2026.
MiniMax’s launch announcement stated that the MiniMax-01 models could be used through MiniMax Chat and the MiniMax Open Platform API. However, current MiniMax text-generation API documentation lists MiniMax-M3, MiniMax-M2.7, MiniMax-M2.7-highspeed, MiniMax-M2.5, MiniMax-M2.5-highspeed, MiniMax-M2.1, and MiniMax-M2 as selectable model options, not MiniMax-Text-01.
Because the current official docs do not confirm MiniMax-Text-01 as a selectable hosted API model ID, this page does not include executable Python or curl examples for MiniMax-Text-01. Production users should verify their MiniMax console, current model list, pricing page, license, and deployment path before implementation.
FAQs
What is MiniMax-Text-01’s context window?
MiniMax official materials state that MiniMax-Text-01 was trained with a 1M-token context and can handle up to 4M tokens during inference, as of June 2026.
How much does MiniMax-Text-01 cost?
MiniMax’s January 2025 launch announcement listed $0.20 per 1M input tokens and $1.10 per 1M output tokens. Current MiniMax pay-as-you-go docs do not list MiniMax-Text-01 pricing as of June 2026.
Does MiniMax-Text-01 have an API?
MiniMax announced Open Platform API access at launch, but current MiniMax text-generation docs do not list MiniMax-Text-01 as a selectable model ID. Open-weight access is verified.
What is MiniMax-Text-01 best used for?
MiniMax-Text-01 is suitable for long-context text analysis, document review, structured extraction research, and open-weight long-context model evaluation, provided users verify licensing, infrastructure, and current access details.
