Fireworks AI Adds Day-0 Support for MiniMax M3 Multimodal Model

Fireworks AI

Jun 13, 2026

Fireworks AI launched day-0 support for MiniMax M3, a multimodal model featuring native image and video input. Powered by MiniMax Sparse Attention, the model delivers 9× faster prefill and 15× faster decode speeds. It is available now on serverless and on-demand endpoints, priced at parity with M2.7 at $0.30 per million input tokens.

MiniMax M3 vs. the frontier
Agentic, coding, and document benchmarks. M3 leads its class and trades blows with the largest closed models.
MiniMax M3 | MiniMax M2.7 | Claude Opus 4.7 | GPT 5.5 | Kimi K2.6 Thinking | Takeaways
SWE-Bench Pro | 59.0 | 56.2 | 64.3 | 58.6 | 58.6 | Surpasses GPT-5.5 and Gemini 3.1 Pro
Terminal-Bench 2.1 | 66.0 | 51.1 | 66.1 | 78.2 | — | On par with Opus 4.7, trails GPT-5.5
SVG-Bench | 63.7 | 48.0 | 62.3 | 58.2 | — | Top score
MCP Atlas | 74.2 | 49.4 | 77.0 | 75.3 | 66.6 | Almost on par with GPT-5.5
Claw-Eval | 74.5 | 49.7 | 71.6 | — | 61.5 | Top score
OmniDocBench | 91.6 | — | 89.3 | 87.5 | — | Top score — MiniMax M3 performance benchmark comparison against frontier models across agentic, coding, and document processing evaluation tasks.

View the full update on fireworks.ai

Fireworks AI

@FireworksAI_HQ1d ago

MiniMax M3 is live on Fireworks. Day-0, fastest endpoint for the MiniMax series. → Top open-weight model on the Artificial Analysis index → 512K context, native image + video input → MSA sparse attention: 9× faster prefill, 15× faster decode → Priced at parity with M2.7 https://t.co/H7xgMwJhpN

581

View on X

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

Keep reading

Fireworks AI hosts MiniMax M3 with 15x faster long context decoding

Fireworks AI is now powering inference for MiniMax M3, a multimodal model featuring a novel sparse attention architecture. The partnership enables 15.6x faster decoding at 1-million-token context, making real-time agentic workflows viable at scale.

MiniMax15h ago

MiniMax Releases M3 Open Weights with Modular Day-Zero Cloud Launch

MiniMax released M3 open weights, a multimodal model featuring a 1-million-token context window and a new sparse attention architecture. Modular is a day-zero launch partner, providing kernel-optimized inference on its cloud. The model, which uses approximately 428 billion total parameters and 23 billion activated parameters, is designed for frontier coding and long-horizon agentic tasks.

Together AI powers MiniMax M3 with 1M context and sparse attention

Together AIJun 3

Together AI powers MiniMax M3 with 1M context and sparse attention

Together AI is now powering inference for MiniMax M3, a multimodal model featuring a 1-million-token context window. The model uses a new sparse attention architecture to process massive datasets with significantly lower computational overhead than previous-generation models.

NVIDIA Launches Free GPU-Accelerated Endpoint for MiniMax M3 Multimodal Model

NVIDIA15h ago

NVIDIA Launches Free GPU-Accelerated Endpoint for MiniMax M3 Multimodal Model

NVIDIA now provides a free GPU-accelerated endpoint for MiniMax M3, a 428B parameter Mixture-of-Experts model. Developers can use the endpoint to test the model's 1-million-token context window and native multimodal reasoning for text, image, and video tasks. The model is available for prototyping and evaluation on the NVIDIA build platform.