Skip to content

PlatformsAMD

AMD introduces the Instinct MI450X GPU and posts MI455X inference results on Helios

AMD Instinct MI450X at a glance

AMD added the Instinct MI450X GPU to its MI400 series on November 12, 2026, a 288 GB HBM4 option for Helios rackscale systems, and reported MI455X inference of 21,333 tokens per second per GPU on DeepSeek-R1 in an MLPerf Offline configuration.

Ask your AI about this story

Use this story in your business: AMD introduces the Instinct MI450X GPU and posts MI455X inference results on Helios (AMD). Ask your AI to explain what happened, how it could affect your business and how to put it to work. It also weighs price, speed, intelligence and the lab behind it. The link to the story is included.

What changed with AMD Instinct MI450X

AMD expanded its Instinct MI400 series on November 12, 2026 with the MI450X, a new GPU option for the AMD Helios rackscale platform that offers 288 GB of HBM4 memory per GPU while keeping the same core compute, memory bandwidth and scale-up and scale-out connectivity as the MI455X. Helios customers can choose either GPU in the same rack design and the same ROCm software stack: the MI450X fits workloads whose memory needs it covers, while the MI455X leaves more headroom for larger models, longer contexts or higher concurrency.

AMD also published new measured inference results for the MI455X on Helios: 21,333 tokens per second per GPU running DeepSeek-R1 in an MLPerf Inference Offline configuration tested by AMD Performance Labs in October 2026, which AMD describes as up to 30% higher throughput than NVIDIA Vera Rubin NVL72's published MLPerf Inference v6.1 result, and up to 20% higher HBM bandwidth than a Vera Rubin GPU on AMD's calculations. The comparison is AMD's own testing against NVIDIA's published figures, not an independent head-to-head.

AMD Instinct MI450X in numbers

Values are copied from the linked source on the date shown. They are not IA Compa measurements.

MI455X throughput, DeepSeek-R1 (MLPerf Offline) · AMD Instinct MI455Xas of · Source for MI455X throughput, DeepSeek-R1 (MLPerf Offline): newsroom.amd.com (opens in a new tab)
21,333 tokens/s per GPU

Why AMD Instinct MI450X matters for businesses

Why it matters: for businesses buying AI capacity, a second memory tier inside the same Helios rack means infrastructure teams can match GPU spend to the models they actually serve instead of buying the largest configuration for every workload, and published MLPerf-format numbers make AMD's inference claims easier to compare when planning capacity outside the NVIDIA ecosystem.

Put it to work: AI services

Get AI News by email

Verified AI stories, each with its official source, on Monday and Thursday mornings. Add breaking news to hear about the biggest ones as soon as we publish them.

Every story links its official source

Free, in English or Spanish

Unsubscribe with one click, any time

What to send

Sources for this story

  1. Official

Last verified: . IA Compa summarizes public announcements for business readers. We are not affiliated with the companies mentioned. Always confirm details, pricing, and availability with the official source.

  • #amd
  • #gpu
  • #instinct
  • #ai-infrastructure
  • #inference

Want this working in your business?

We turn new AI capabilities into agents, voice receptionists, and automations you can actually use.