mailwal

SambaNova Lightning Digest: MiniMax M3.1 Flash lands on SambaCloud

SambaNova

This month, a fast, low-cost model for agents arrives on SambaCloud: MiniMax M3.1 Flash, running at over 200 tokens per second. We also explain what an AI factory is and why it matters once AI moves into production. Then we turn to the events: the Supercomputing Conference (SC26) in Chicago from November 15–20, and NeurIPS from December 6-12, where we'll have a booth. Read on, and come say hello.

NEWS YOU CAN USE
MiniMax M3.1 Flash Lands on SambaCloud

A new option for long-running coding and tool-using agents is available on SambaCloud at over 200 tokens/s. If you call MiniMax M3 today, note that requests are being rerouted to M3.1 Flash; update your model string to MiniMax-M3.1-Flash.

SambaNova and Oracle Bring Fast Inference to OCI

SambaNova and Oracle are teaming up to deploy SambaNova's SN50 inference chips in Oracle Cloud Infrastructure data centers. Oracle will offer the capacity to AI model builders as agentic AI pushes demand for fast, cost-efficient inference. SambaNova claims up to 5X the speed of competing chips at 3X lower total cost of ownership.

Prompt Caching for MiniMax 3

Repeated, stable prefixes in agent prompts can now be served from cache on SambaCloud with no code changes. The post explains how to arrange context for cache hits, inspect cached_tokens, and reduce time to first token and input cost. This same feature is inherited into MiniMax-M3.1-Flash too!

What is an AI Factory?

This practical explainer walks through the infrastructure, models, data, and governance needed to move from AI experiments to production—and why an inference-focused token factory cares about throughput, watts, and cost to serve.

Join Us At Events Online & Around the World

SC26 – Supercomputing Conference

The HPC community meets in Chicago, November 15–20, 2026. Explore the technical program for conversations about large-scale computing and AI infrastructure. Stop by our booth to talk shop.

NeurIPS 2026 - Neural Information Processing Systems

The fortieth NeurIPS runs December 6–12, 2026, in Sydney, with satellite events in Atlanta and Paris on December 9–13. SambaNova will be there with a booth, so stop by to talk shop and see what's running on SambaCloud.

Stay Connected
Stay connected, keep innovating, and let’s shape the future of AI together! For more information, follow us on X and Linkedin to stay up to date with the latest announcements on SambaCloud. Also, join the SambaNova Community for the latest information and showcase what you are building!
2025_Sambanova_Logo2_1C-White_RGB

2460 N First St, Suite #100, San Jose,CA 95131 USA

Copyright © 2026 SambaNova, Inc. All rights reserved.