Wolfram Ravenwolf
@WolframRvnwlf
AI Engineer by title. AI Evangelist by calling. AI Evaluator by obsession. Evaluates LLMs for breakfast, preaches AI usefulness all day long at @ellamindAI. 😎
Here's an updated version of my guide "How to use Qwen3-Coder (or any other LLM) with Claude Code": gist.github.com/WolframRavenwo… It now includes extra helpful tips - and building your own image is no longer necessary.
I've expected the new Qwen reasoning model to be on par with the closed source SOTA. Their non-reasoning model already topped everything in open source, according to their benchmarks and my own. China isn't catching up, they already caught up, and they're sharing this freely! 👏…
Wow the new qwen reasoner at only 232B params is as good as the top closed frontier lab models Big day for OS
If you missed the live show (Shame! Shame! Shame! *bell ringing*), repent and catch up here:
🔥 WHAT a WEEK! Qwen-mas in July, America's AI Action Plan, Math Olympiad gold for AIs & we coded a 3D Tetris live! ThursdAI is on fire! Here's your high-signal recap 👋👇 #ThursdAI
🚨 LIVE - 📆 ThursdAI - Qwen3-Coder & new A22B, 🇺🇸 AI action plan, LLMs win IMO, Sapient HRM & more AI news x.com/i/broadcasts/1…
I've just completed my initial MMLU-Pro CS benchmark runs for Qwen3-235B-A22B-Instruct-2507. Based on both the scores and my personal experience with the model, I can confirm: this is the current reigning champion of open-source/weights LLMs! 👏 Amazing work, @Alibaba_Qwen team!
New Qwen3 release! If this is a "small update" (look at those scores!), then I can't wait for what they consider "bigger things". 😅 And it also marks a shift away from hybrid thinking models. I appreciated the all-in-one approach, but if separate models prove superior, then…
I'm glad it's available (thanks @sama), but those long cycles are harsh! I'd prefer 1 or 2 uses per day rather than 40 in a month because I can't predict my needs that far in advance. I'll probably try to save them until the end of the cycle. Even better, let me pay per use once…
I feel like I'm stating the obvious, but: your AI can learn if you equip it with the right tools and prompts to read and write important information and lessons learned during and after each session. Without such tools, using AI is like operating a workstation without peripherals…
While others delay their models, don't release the weights at all or talk about going closed source, others - like @Alibaba_Qwen - keep on giving. And not just some scraps, but their actual SOTA open source/weights models. 👏 Join us on ThursdAI when we discuss this in detail.
💥 BREAKING: @Alibaba_Qwen just dropped the worlds leading coding model - a 480B Qwen3 Coder with 35B active parameters and a huge context window! This non-reasoning coder is getting near SOTA at SWE-bench, 68.7 on BFCL (function calling) and 61.8 on Aider! 🧵
New Qwen3 release! If this is a "small update" (look at those scores!), then I can't wait for what they consider "bigger things". 😅 And it also marks a shift away from hybrid thinking models. I appreciated the all-in-one approach, but if separate models prove superior, then…
Bye Qwen3-235B-A22B, hello Qwen3-235B-A22B-2507! After talking with the community and thinking it through, we decided to stop using hybrid thinking mode. Instead, we’ll train Instruct and Thinking models separately so we can get the best quality possible. Today, we’re releasing…
So what's the best web search MCP server? I'd prefer something free and open source for the search backend as well (ideally run locally), so looked for SearXNG compatibility, but the MCP servers I found weren't all that popular. What are you guys using?
Coolest demo I've seen in a while! Tried it myself, works in real time, showing off what AI-augmented (and altered) reality will look like in the near future when we have our AI/AR glasses. I've anticipated and awaited this tech for many years, and now it's finally within reach.
Introducing MirageLSD: The First Live-Stream Diffusion (LSD) AI Model Input any video stream, from a camera or video chat to a computer screen or game, and transform it into any world you desire, in real-time (<40ms latency). Here’s how it works (w/ demo you can use!):
Important advice - the better you understand the tools you're using, the less you risk hurting yourself while using them!
ChatGPT psychosis is rare if you know how it works. “How GPT works” should be a required subject. Primary lesson: a GPT is a function of its data. It can uncover interesting facts and make fascinating correlations. But it will not discover non-governmental demons. Anything…
😂
"[video game] as a community theater production" may be one of the most delightful Veo 3 Fast prompts Please enjoy, in order: GTA, Pokemon, Mario Kart, The Witcher 3, Stardew Valley, Tetris, Mortal Kombat, The Sims, & Death Stranding(!) Yes, the whole prompt was the one above.