J. Jubran retweeted
ZHA is growing up!
We’ve scaled @ZyphraAI’s subquadratic attention method - ZHA - to leading open source multi-trillion parameter models across Nemotron, Qwen, Mimo, and GPT-OSS!
ZHA preserves reasoning and agentic performance while delivering 9-60x long-context speedups at 10M tokens.
J. Jubran retweeted
We scaled Zyphra Hierarchical Attention to 2.4 trillion parameters. Across four open-weight models, we kept the backbones frozen and preserved performance on most reasoning, coding, and long-context benchmarks.
We’ve scaled @ZyphraAI’s subquadratic attention method - ZHA - to leading open source multi-trillion parameter models across Nemotron, Qwen, Mimo, and GPT-OSS!
ZHA preserves reasoning and agentic performance while delivering 9-60x long-context speedups at 10M tokens.
GLM-5.3-Flash (FP4) is on Zyphra Inference at 25% below our FP8 price, served with the Zyphra Optimization Engine (ZOE) on @AMD Instinct MI300X.
With ZOE, p90 generation is 9.5x faster per user and p90 TTFT is 48x lower than stock SGLang.
Try it at: cloud.zyphra.com
@ZyphraAI is on the @ArtificialAnlys leaderboard for GLM-5.3-Flash throughput. Optimized on @amd MI300X with Zyphra’s Optimization Engine (ZOE).
Try it on Zyphra Inference: cloud.zyphra.com
$/1M tokens:
- Input: $0.15
- Cached: $0.03
- Output: $0.25
GLM-5.3-Flash from @zai_org is on Zyphra Inference, served with the Zyphra Optimization Engine (ZOE) on @AMD Instinct MI300X.
With ZOE, p90 generation is 3.1x faster per user and p90 time to first token is 10.7x lower than stock SGLang.
Try it at: cloud.zyphra.com
Zyphra presents the second wave of inference optimizations for @AMD Instinct MI355X GPUs, powered by Zyphra Optimization Engine (ZOE).
Our latest engineering blog details the kernel work behind faster long-context serving for frontier open models. 🧵
Anthropic at $965B edging out OpenAI at $852B is the clearest signal yet that private AI valuations have entered pure narrative territory.
The revenue ramps are real, the talent wars are brutal, and Databricks at $188B shows the enterprise layer still has gravity. But these numbers only work if the eventual public markets keep buying the story at the same multiple. Paper crowns don’t pay the GPU bills.
J. Jubran retweeted
Everyone is focused on record AI capex but the bigger story is that ~$2.3T of cloud backlog is already under contract and waiting to be deployed.
The demand is already there but the constraint is that the data centers, power and chips simply don't exist yet to fulfill it.
Q2 Cloud Backlog:
• $MSFT ~$678B
• $ORCL ~$638B
• $GOOGL ~$514B
• $AMZN ~$496B
J. Jubran retweeted
Honored that IBM CEO, Arvind Krishna, is featured in the top 10 CEOs of 2022 on @Comparably’s Best CEOs list. Congratulations, @ArvindKrishna: ibm.co/3GnvV8R
I don’t understand it but watching people play Pokémon go in Seattle today made me a fan. #PokemonGOFestSeattle
1
Why? Economic toll in Texas worsens as trucks remain stopped at Mexico border - The Washington Post @GregAbbott_TX apple.news/AZx2Ou98QTciI7fOe…
1
J. Jubran retweeted
Since the Russian invasion of Ukraine started two weeks ago, China has tried to keep its head down.
But Beijing is now deeply exposed to turmoil. Here is why 👇on.ft.com/3hVMZFY
J. Jubran retweeted
Video for High Performance Computing (HPC) on the Cloud with Spectrum LSF, Spectrum Symphony and Slurm covering business value, architecture and demo. Please share
youtu.be/ozhbYml-WOw
@HPCwire @HPC_Guru @ThisWeekinHPC @intelhpc @insideHPC @Supercomputing #HPC #CloudComputing
World’s Largest Wildlife Crossing Is Now Under Way in L.A. apple.news/ALeBSRv3LRjyD73_w…
1
J. Jubran retweeted
A new @IBMIBV global study on cloud transformation reveals that hybrid cloud is the dominant IT architecture. I write about this key finding and others from the study in my new blog post. linkedin.com/pulse/whats-hyb…
J. Jubran retweeted
Join our IBM Cloud virtual event, Cloud without Compromise: Reduce risk. Innovate anywhere. Co-hosted by Reggie Watts, musician, comedian and beatboxer along with speakers from leading organizations. ibm.co/3gKOSFE

























