📣 Announcing the world’s fastest AI platform SambaNova Cloud runs Llama 3.1 405B @ 132t/s at full precision ✅ Llama 3.1 405B @ 132 tokens/sec ✅ Llama 3.1 70B @ 570 tokens/sec ✅ 10X Faster Inference than GPUs Start developing #FastAI ➡️ cloud.sambanova.ai @AIatMeta

Do you want to test what we have been building at @SambaNovaAI over the last few years? You can now play around with our API, it’s OpenAI-compatible, free and fast! (1000 tokens/second on Llama 3.1 8B at full precision; 4X faster than GPUs) sambanova.ai/fast-api?api_re…

Artificial Analysis has independently benchmarked @SambaNovaAI's custom AI chips at 1,084 tokens/s on Llama 3 Instruct (8B)! 🏁 This is the fastest output speed we have benchmarked to date and >8 times faster than the median output speed across API providers of @Meta's Llama 3 Instruct (8B) we benchmark. SambaNova currently does not yet publicly offer a serverless API but you can try out their system via their chat interface (see below tweet). SambaNova is not yet listed on the Artificial Analysis leaderboard but we understand API services using SambaNova chips will be available in the near future and we look forward to initiating full coverage. SambaNova’s custom SN40L RDU chips are their fourth generation design and are built on TSMC’s 5nm process. They are reported as having the potential to scale to serve much larger models than Llama 3 Instruct (8B) - Llama 3 400B 👀. Artificial Analysis has also verified that Llama 3 Instruct (8B) on Samba-1 Turbo achieves quality scores in-line with full FP16 precision by testing an MMLU-based benchmark.

Feels great to see this release finally out.

Java's virtual threads are awesome 🥳 I thought long about if I should run a preview feature of an early access JDK-19 in production😅 But it is the only solution I found that solves this problem at the right level without creating accidental complexity clojureverse.org/t/strategy-…

SL be like: The bus is late, but we’ll solve this by teleporting it. So you’ll arrive 24 minutes ahead of schedule.

It seems my laptop is now 10 years old. New personal best.

Awaiting a package. US to Stockholm. ETA today 12:00. Current location Memphis. Current status according to Fedex: On time. Will they make it?

Targeted to JDK 19: JEP 428: Structured Concurrency (Incubator): openjdk.java.net/jeps/428 #jdk19 #openjdk #java

People (companies) need to stop using, and benchmarking, the experimental JDK 11 version of ZGC. It was the first version we released, and we haven't backported any bug fixes or improvements to it. All fixes and enhancements have gone in to the later JDK releases.

Concurrent Thread-stack Processing in the Z Garbage Collector // via @eosterlund_fisk #OpenJDK #ZGC #Java inside.java/2022/05/31/zgc-c…

Awesome to have friends in different fields.

Virtual threads are amazing, but only half of Loom's play. The other half is making use of quasi infinite threads with a new concurrency programming model: Structured Concurrency. JEP 428 proposes that - will it make it into #JDK19? 🤞 openjdk.java.net/jeps/428

📢 Heads-up - JEP 425 Virtual Threads (preview) Testing #Java #Java19 #JDK19 #OpenJDK #ProjectLoom inside.java/2022/05/16/quali…

Proposed to target JDK 19: JEP 405: Record Patterns (Preview): openjdk.java.net/jeps/405 #jdk19 #openjdk #java

On the 8th of March, we got one of Norway's biggest CS players Jkaem to play under a female alias to raise awareness about one of the dark corners in the gaming world. #dontchangeyourname

JDK 8 to JDK 18 in Garbage Collection: 10 Releases, 2000 Enhancements #Java #OpenJDK inside.java/2022/05/02/odl-j…

mail.openjdk.java.net/piperm… Async await and Kotlin coroutines are great, but even better is VM support for abundant threads that make blocking cheap. microsoft.com/en-us/research…

If you want to read more about the fascinating 10-year-old bug that could cause minutes long application stalls in concurrent GCs, see the PR: github.com/openjdk/jdk/pull/…

ZGC | What's new in JDK 18 malloc.se/blog/zgc-jdk18