MiniMax M3.1 Flash Launches After Space Bunny AI Model’s Stealth Run

Chinese artificial intelligence company MiniMax has introduced MiniMax M3.1 Flash, a new AI model designed to improve coding, autonomous task execution and cybersecurity-related capabilities. Announced on October 19, 2026, the release also reveals that the mysterious Space Bunny model, which gained attention among developers in recent weeks, was an early version of MiniMax’s latest technology.

According to Reuters, MiniMax claims the model delivers significant performance improvements while reducing usage costs by approximately 70% compared with its earlier offering.

The company also reported that Space Bunny processed more than 200 trillion tokens during its anonymous testing period.

What Is MiniMax M3.1 Flash?

MiniMax M3.1 Flash is an AI model designed for developers and applications requiring efficient reasoning, programming assistance and automated workflows.

The model supports multimodal inputs and offers a context window of up to one million tokens, allowing it to process large amounts of information within a single interaction.

MiniMax says its latest release focuses on three areas:

• Agentic tasks: Supporting AI systems that complete multi-step activities and automated workflows.
• Programming: Assisting with code generation, debugging and software development.
• Cybersecurity: Improving capabilities for security-related analysis and technical tasks.

The company describes these improvements as generational advances, although independent testing is needed to establish how consistently they perform across real-world applications.

Space Bunny Revealed as an Early MiniMax AI Model

Before the official launch, an anonymous model called Space Bunny Alpha appeared on OpenRouter on September 23, 2026.

Its developer was initially undisclosed, prompting speculation within the AI community.

By September 24, the model had reportedly reached the top of OpenCode’s daily token-usage rankings.

MiniMax later confirmed that Space Bunny was an early version of M3.1 Flash.

According to the company, the anonymous model processed more than 200 trillion tokens during its testing period.

OpenCode also reported that Space Bunny held the leading position on its platform for 12 consecutive days and handled approximately 106 trillion tokens.

These figures reflect reported platform usage rather than independently verified measures of model quality.

MiniMax M3.1 Flash API Pricing and Availability

One of the main features of MiniMax M3.1 Flash is its reported lower operating cost.

MiniMax says the model costs approximately one-third as much as its earlier M3 model while delivering roughly three times the inference speed.

API usage	Price per million tokens
Input tokens	$0.15
Output tokens	$0.60
Cached input reads	$0.015

API prices are based on MiniMax’s reported launch pricing and may change.

The model is available through the MiniMax API, M Plan subscription and MiniMax Code desktop application.

OpenRouter and OpenCode also announced temporary free-access periods following the launch. OpenRouter’s promotional access was scheduled for 72 hours, while OpenCode announced a three-day offer.

MiniMax plans to release the model’s open weights on October 23, although its parameter count and licensing terms have not yet been disclosed.

MiniMax M3.1 Flash Benchmark Performance

MiniMax published benchmark results comparing its new model with competing AI systems.

The company reported improvements in automation-related tasks while acknowledging that competing models performed better in certain coding and visual-reasoning evaluations.

Benchmark	MiniMax M3.1 Flash	Competing model	Competitor score
Automation-Bench	62.2%	DeepSeek-V4.1-Flash-1017	60.5%
Automation-Bench	62.2%	Claude Opus 5	50.3%
DeepSWE	73.7%	DeepSeek-V4.1-Flash-1017	77.0%
BabyVision	91.0%	Claude Opus 5	94.1%

Results are based on MiniMax’s published comparisons. They should not be interpreted as independently verified benchmark rankings.

The figures suggest that M3.1 Flash may be competitive in automation tasks, although performance varies depending on the evaluation.

Faster AI Responses During the Space Bunny Testing Period

MiniMax reportedly improved Space Bunny’s performance while the model was available anonymously.

According to OpenRouter, several operational measurements improved during the testing period.

Performance metric	Earlier	Later
Time to first token	1.95 seconds	1.22 seconds
Generation throughput	70 tokens/second	100 tokens/second
Error rate	1.4%	0.7%

These changes indicate faster initial responses, improved generation throughput and fewer reported errors under the measured conditions.

Actual performance may vary depending on workload, infrastructure and API demand.

MiniMax Plans a More Powerful Pro Model

MiniMax has also confirmed plans to introduce a higher-end Pro model.

While the company has not announced a release date, pricing or detailed specifications, it says the upcoming version will focus on more advanced capabilities.

The Flash release reflects MiniMax’s strategy of balancing model performance with speed and operating costs.

For developers, lower API pricing and a large context window could make the model an option worth evaluating for coding assistants, AI agents and other applications.

Frequently Asked Questions

1. What is MiniMax M3.1 Flash?

MiniMax M3.1 Flash is an AI model designed for programming, agentic workflows and other technical applications, with multimodal input support and a one-million-token context window.

2. Is Space Bunny the same as MiniMax M3.1 Flash?

MiniMax confirmed that the anonymous Space Bunny model was an early version of M3.1 Flash.

3. How much does MiniMax M3.1 Flash cost?

Reported API pricing starts at $0.15 per million input tokens and $0.60 per million output tokens, with cached reads priced at $0.015 per million tokens.

4. Is MiniMax M3.1 Flash open source?

MiniMax announced plans to release the model’s open weights on October 23, 2026. The licence has not yet been disclosed, so open-source status cannot be confirmed.

5. Where can developers access MiniMax M3.1 Flash?

The model is available through the MiniMax API, M Plan and MiniMax Code, with additional access offered through platforms including OpenRouter and OpenCode.

Conclusion

The launch of MiniMax M3.1 Flash brings an end to speculation surrounding the anonymous Space Bunny model and introduces a lower-cost AI option for developers.

With reported improvements in inference speed, automation capabilities and API pricing, the model could appeal to teams building AI-powered software and automated workflows.

However, independent performance evaluations and details about the planned open-weight release will be important in assessing its broader capabilities and accessibility.