Releases·Model Release
MiniMax launches M3.1 Flash Preview for Token Plan subscribers
MiniMax announced that MiniMax-M3.1 Flash Preview is now live on the Token Plan. The company positions it as a faster, lighter model for teams running high-volume, latency-sensitive workloads. It is available under existing Token Plan subscriptions with no extra setup required.
469 engagement1 sourceSun, Sep 27, 2026, 05:06 PM
AI Overview
- MiniMax-M3.1 Flash Preview is live for Token Plan subscribers with no extra setup required, per MiniMax's announcement.
- MiniMax positions M3.1 Flash for teams running high-volume, latency-sensitive workloads.
- MiniMax bills M3.1 Flash as faster and lighter, targeting high-throughput use cases.
Related
Releases·Model Release
MiniMax releases open-weight H3 video model with native stereo audio and multimodal reference control
MiniMax promoted its open-weight H3 video generation model, built with native stereo audio and multimodal reference control. The company says open-source community adoption is making the capability faster and easier to build on.
MiniMax promoted its open-weight H3 video generation model, built with native stereo audio and multimodal reference control. The company says open-source community adoption is making the capability faster and easier to build on.
983 engagement1 sourcexMon, Sep 14, 2026, 06:14 AM
Developments·Benchmark
MiniMax-M3 tops agent benchmark at $0.018 per task
MiniMax-M3, a frontier lab model, completed a real-world agent benchmark task—spinning up an inbox and sending a business email end-to-end—for $0.018, the lowest cost on the board. The result highlights the model's efficiency for practical agentic workflows.
MiniMax-M3, a frontier lab model, completed a real-world agent benchmark task—spinning up an inbox and sending a business email end-to-end—for $0.018, the lowest cost on the board. The result highlights the model's efficiency for practical agentic workflows.
73 engagement1 sourcexTue, Aug 25, 2026, 11:51 PM
News·Company News
MiniMax's Morgan Suo to present "The Hidden Costs of a Faster Model" at AI Engineer NYC
MiniMax announced it will speak at the AI Engineer NYC conference. Morgan Suo, Head of US Business Development, will present 'The Hidden Costs of a Faster Model,' covering how quantization, speculative decoding, and reasoning controls affect a model's cost, speed, and quality. The talk targets practitioners weighing inference-time optimizations against quality tradeoffs.
MiniMax announced it will speak at the AI Engineer NYC conference. Morgan Suo, Head of US Business Development, will present 'The Hidden Costs of a Faster Model,' covering how quantization, speculative decoding, and reasoning controls affect a model's cost, speed, and quality. The talk targets practitioners weighing inference-time optimizations against quality tradeoffs.
33 engagement1 sourcexFri, Oct 2, 2026, 01:27 AM
Social·Community
MiniMax M3 weights use restrictive license, MSA kernel is MIT
A Reddit post clarifies that MiniMax's M3 model weights are released under the MiniMax Community License, not open source, while the MSA (MiniMax Sparse Attention) kernel is MIT licensed. The distinction matters for developers evaluating M3 for long-context coding and agent workflows who may want to ship products.
A Reddit post clarifies that MiniMax's M3 model weights are released under the MiniMax Community License, not open source, while the MSA (MiniMax Sparse Attention) kernel is MIT licensed. The distinction matters for developers evaluating M3 for long-context coding and agent workflows who may want to ship products.
4 engagement1 sourceredditWed, Jul 22, 2026, 06:36 PM
Products·launched
Open-source real-time video generation built on vLLM-Omni and FastH3
An open-source real-time video generation release from MiniMax, built on vLLM-Omni and Hao AI Lab's FastH3 with NVIDIA hardware support. Real-time generation enables interactive video applications, and the open baseline lets the community improve it. It targets developers and researchers wanting a public, extensible foundation for streaming video generation.
An open-source real-time video generation release from MiniMax, built on vLLM-Omni and Hao AI Lab's FastH3 with NVIDIA hardware support. Real-time generation enables interactive video applications, and the open baseline lets the community improve it. It targets developers and researchers wanting a public, extensible foundation for streaming video generation.
182 engagement1 sourcexTue, Sep 1, 2026, 11:11 PM