I created a petition to Ask OpenAI to Stop Turning Codex Quota Resets Into Theatre - Sign the Petition! c.org/gYjTj9Zw4G?v=2

DeepSeek V4.1 Flash on Strix Halo. I spent the last week adding support to DwarfStar (@antirez ) and testing two ways to run it: SSD streaming on one 128 GB machine, or distributing the model across two nodes over TCP or RoCE. piped.video/1DaMkTuiCEQ

I’m not exaggerating: Codex usage feels completely out of line right now. I burned through my entire ChatGPT Plus quota in roughly an hour of real work. That is NOT acceptable. Something here feels wrong.

The Astra honeymoon could not last, we knew it. @OpenAI silently slashed all the usage quotas, as of the past few days Codex is unusable even on the most expensive tier, a few hours burn through the entire quota.

This is amazing work, can't wait to cover it on my channel.

First time crossing over 6,000 tok/s prefill on my dual R9700's on Qwen3.8 27b NVFP4. I still remember reaching 3,000, 4,000 and then over 5,000 on these dense models. Everytime I keep thinking it's at the peak, but they keep getting faster. @AIatAMD @RadeonPRO #AMD #R9700 #localLLM #localAI

Recently, we worked on something special: pushing ROCm to beat Vulkan on Strix Halo. Lucebox engine (ROCm) and Llama.cpp (Vulkan) ran on the same Strix Halo on the same day with the same prompt. With the DSpark drafter Lucebox ROCm decodes at 41.9 tok/s at 8K and 38.0 tok/s at 123K prompt tokens. That is 28% to 47% faster than llama.cpp and Vulkan v0.7.5 running the same drafter and its prefill is 17% to 47% faster 🧵

github.com/Nathanw1014/strix… v0.7.6 is out, built in parallel with halo-box (PR coming) - Prompt processing +16% to +31% on standard models tested - Flash-Next PP flat with depth now, +34% at 32k - MTP speculative decoding ~9% faster on 27B with slightly better acceptance

DwarfStar running DeepSeek v4.1 Flash on a 128GB M5 Max. I didn't expect with SSD streaming it could be so fast. Recent SSD streaming changes to retain the right experts surely helped, but also maybe DS4.1 uses the same experts more. Will push online when ready QA > ASAP.

We have a positive update today on LPCAMM2 pricing! We were able to source a limited quantity at lower cost, and are retroactively reducing prices for orders that shipped at higher price, as well as offering the lower price to orders through Batch 10.

Big open source news for Strix Halo owners. 🚀 For the last two weeks we have been working on a community fork of llama.cpp to make it easier for the community to use a stable version and avoid having to switch forks on a weekly basis. Try it out here: github.com/halo-box/strix-ll… Well done @HiGaetan, @NathanW1014 and others, let's continue this work to push the performance of local inference, fun working with you guys! For users with other hardware we have a more mainline compatible version: github.com/halo-box/llama.cp…

Threadripper Halo station: the world's most powerful workstation @AMD

Here we go!

I am getting Google Antigravity vibes here, it's a shame as I was really liking OpenAI's models. but they really messed up the past 12 hours or so with capacity issues and now ChatGPT is down.

Excited to announce that Lucebox engine now serves Qwen3.8-27B on a single AMD Radeon AI PRO R9700 (32 GB, RDNA4) with the DFlash2 block-diffusion drafter from z-lab: - up to 227 tok/s on code - 208 tok/s HumanEval average - math at 133 tok/s on an off-the-shelf quant that matches an 8-bit reference on HumanEval and GSM8K. And it is lossless: greedy verification commits only tokens the model itself would produce, so the drafter changes how fast you get the answer, never what the answer is. Full breakdown in the article below👇

When @TfL says "When you see something unusual..." I always think of calling them if I see a journey without disruption. But these hasn't happened yet...

It's an excellent day for running open weights models locally! Qwen3.8-Flash-Next fits in its entirety at 4-bit on a 128GB Framework Desktop, and at 8-bit on the upcoming 192GB Framework Desktop.

The struggle is real at @github , they just can't get the actions back up. BUT, this is a free service for the OSS community, thank you to all the engineers working hard there. The frustration is mostly because this is happening at bad times, causing a lot of issues.