FHILY👑
@Oluwaphilemon1 • 7,978 subscribers
Defi + Content Creator + Entertainment + Claude News + AI
Shorts
Videos
Сlaude Design is so cracked at marketing. 🤯 and anyone can achieve the same results. Here's the prompt ↓
FHILY👑219,605 görüntüleme • 6 ay önce
Qwen3.8-Flash-Next is still going strong at 364.7K tokens of context on an M5 Max. And this isn’t just a static long-context test. The model was reasoning about how to speed up its own workflow while using tools, and the tool calls kept working without misses. Setup: • Qwen3.8-Flash-Next • M5 Max • 128GB unified memory • MLX-Serve PR #363 • OpenCode 2 • 364.7K context The interesting part isn’t simply getting hundreds of thousands of tokens into memory. It’s what happens once the context gets this large. Long-context inference usually comes with a painful tradeoff. As the KV cache grows, memory pressure increases and generation can slow down. But this setup is still pushing through 364K tokens while maintaining a usable agent workflow. The model can reason, call tools, inspect results, continue working, and keep the session moving. And the tool calls reportedly haven’t missed so far. That’s important for agentic coding. A huge context window is only useful if the model can actually operate reliably inside it. A 400K-token context that constantly breaks tool calls isn’t very useful. A 364K session that can keep reasoning and executing tools is a different story. And the test isn’t finished yet. The current run is approaching 400K tokens, with the expectation that it can keep going. This is also another interesting example of why Apple Silicon keeps showing up in local LLM experiments. The M5 Max’s unified memory gives a large model and its growing KV cache access to one shared memory pool. With MLX-Serve continuing to improve, these machines are becoming surprisingly capable long-context inference boxes. The bigger takeaway: Context length is becoming a workload, not just a model specification. Running a model at 256K is one thing. Keeping an agent alive at 300K+ while it reasons and uses tools is much more interesting. And Qwen3.8-Flash-Next is showing that this can be pushed surprisingly far on a single 128GB Mac. 364.7K and counting. Next stop: 400K.
FHILY👑39,982 görüntüleme • 1 ay önce
125B on a single RTX 4090 at 25 tok/s is actually insane Qwen3.8-Flash-Next + MTP is hitting: 80K context 25.35 tok/s decode 471 tok/s prefill On one consumer GPU. Yes, it’s an MoE, so you’re not running all 125B parameters for every token. But still, getting a model this large to run this fast on a 4090 is wild. MTP is doing a lot of the heavy lifting here too. Instead of generating one token at a time, it can speculate on multiple tokens and verify them together. And 25 tok/s is the important number.
FHILY👑40,049 görüntüleme • 1 ay önce
Qwen3.8-27B spent 21 days trying to build its own CUDA inference engine on a single RTX 3090. Mostly unsupervised. No human-written CUDA. Only around 12 human nudges across the entire experiment. The model wasn’t just given one giant prompt and left alone either. DeepSeek Harness managed the whole operation: Subagents. Roles. Handoffs. Context management. Compaction. HyperQwen served Qwen3.8-27B through vLLM while the agents worked on the inference engine. And the project accumulated some serious numbers over those three weeks: 180 subagents 230M tokens processed 1.7B cache-read tokens 699 context compactions 83 hours spent on context compaction The final engine didn’t beat llama.cpp. That’s important. HyperQwen managed roughly 250 prefill tok/s compared with around 700 tok/s for llama.cpp in the reported comparison. So this isn’t a story about an AI suddenly inventing a better CUDA runtime. It’s a story about how far you can push a local 27B model when you give it an engineering environment, persistent state and enough time. The funniest part was what happened when the agent started testing its own work. The RTX 3090 has 24GB of VRAM, and the same GPU needed to run both Qwen3.8-27B through vLLM and the CUDA engine being tested. That meant the workflow had to repeatedly: Stop vLLM. Run the benchmark. Restart vLLM. Continue the project. Then one subagent ignored the handoff protocol and killed vLLM while the orchestrator was still running. It basically shut down its own brain. A genuine AI suicide loop. And yet the larger system recovered and continued the project. That part is more interesting to me than whether the resulting inference engine beat llama.cpp. A quantized Qwen3.8-27B model maintained a real software engineering project for three weeks on consumer hardware. It spawned 180 subagents. Processed 230M tokens. Compacted its own context 699 times. And required only about a dozen human interventions. That’s a very different kind of benchmark. We’re starting to move from: “Can the model write code?” to: “Can the model maintain a software project long enough to actually get somewhere?” The answer here wasn’t “it built a better inference engine.” It didn’t. But it did demonstrate that a local Qwen3.8-27B can participate in a surprisingly long-running engineering workflow without a human sitting there writing the implementation. One clarification because the attached video can be misleading: The video is NOT the CUDA engine produced during the 21-day experiment. It compares HyperQwen, the inference backend used to run Qwen3.8-27B during the experiment, against stock vLLM on the same RTX 3090 using the same prompts. The 21-day experiment and the video comparison are separate things. The engine lost the speed race. The agent endurance experiment is still pretty wild.
FHILY👑25,177 görüntüleme • 28 gün önce
Qwen3.8-27B running at full BF16 on a free Kaggle TPU is kind of ridiculous. No quantization. No tiny context window. No expensive GPU instance. Just Qwen3.8-27B running on a Kaggle TPU v5e-8. The reported numbers: ~130 tok/s decode ~10,000 tok/s prefill 262K context That prefill number is especially wild. You can throw a huge amount of code or context at the model and ingest it extremely quickly, while still getting around 130 tokens per second during generation. And because it’s running in full BF16, you’re not relying on an aggressive quant just to make the model fit. But the really interesting part isn’t even the raw throughput. You can expose it as an OpenAI-compatible endpoint. That means you can plug the model into tools that already understand OpenAI-style APIs. Claude Code. Codex. OpenCode. And other compatible clients. So the workflow becomes pretty simple: Spin up the Qwen3.8-27B endpoint on Kaggle. Point your coding tool at the API. And suddenly you have a 27B coding model sitting behind the same interface you’d normally use for hosted models. The 262K context is also a huge deal for agentic coding. Large repositories can fit into a single context. Long conversations don’t need to be constantly trimmed. And tools can feed much more information back to the model without hitting a tiny context ceiling. The fact that this can be built around a free TPU environment is what makes this especially interesting. We’re getting to a point where experimenting with serious open models doesn’t always require owning a $2,000 GPU or paying for a large cloud instance. Free compute + open weights + an OpenAI-compatible API + existing coding agents. That’s a pretty powerful combination. Qwen3.8-27B is already an interesting model. Running the full BF16 version at ~130 tok/s with 262K context on free Kaggle TPU compute makes it a lot more interesting.
FHILY👑35,910 görüntüleme • 1 ay önce
Qwen3.8-Flash-Next just got a massive speed boost on my old hardware. This guy started with roughly 11 tok/s. After tuning the inference stack, I’m now getting around 24–28 tok/s. That’s roughly a 2× speedup without changing the GPU. The setup: → RTX 3090 24GB → 64GB DDR4 system RAM → Qwen3.8-Flash-Next → Cafe-llama.cpp The interesting part is that Cafe-llama.cpp isn’t just another llama.cpp fork. The optimizations are making a real difference for this particular architecture, especially when you’re dealing with a model that relies heavily on system memory and offloading. Going from 11 tok/s to nearly 30 tok/s completely changes how practical the model feels for local use. And now I’m curious what happens with DDR5. If memory bandwidth becomes another bottleneck, a faster RAM subsystem could squeeze even more performance out of the same 3090. This is why local AI optimization is getting so interesting. Sometimes you don’t need a new GPU. You just need a better inference stack. 11 tok/s → 24–28 tok/s.
FHILY👑23,243 görüntüleme • 1 ay önce
changed how I design with AI. I show how to build faster by adding a design system directly into the prompt, importing any website URL to get a strong starting point, then polishing everything in my own style. I also mix it with components and skills to keep the design consistent and more unique. GPT-5.5 cloning websites is honestly crazy good. The result can get very close to the original, then you can refine it and make it your own. Full breakdown in the comments.
FHILY👑57,007 görüntüleme • 4 ay önce
Qwen3.8 Flash Next is starting to look ridiculous on Apple Silicon. I’m running the 4-bit MTP build locally on an M3 Ultra Studio, and the latest OMP run hit: 97.1 tok/s decode 23K context ~1,131 tok/s uncached prompt processing That first number is the one that caught my attention. Nearly 100 tokens per second from a local Qwen3.8 Flash Next setup is already fast enough that the usual “local models are slow” argument starts feeling pretty outdated. And the prompt processing speed is even crazier. Over 1,100 tok/s on an uncached prompt means the model can chew through a large amount of context before generation even starts. The setup matters here. This isn’t just Qwen3.8 Flash Next running untouched. It’s a 4-bit quantized build with MTP, and the inference stack is clearly doing a lot of work behind the scenes to make the hardware perform like this. There’s already a PR open for the implementation on oMLX, so this isn’t just a one-off local experiment either. If these optimizations make their way into the broader MLX ecosystem, running large models locally on Apple Silicon gets even more interesting. The other thing I like about numbers like this is that they put the focus back on the entire inference stack. Model size is one variable. Quantization is another. Then you have MTP, KV cache configuration, runtime optimizations and the hardware itself. Change the recipe and the same model can feel completely different. Qwen3.8 Flash Next at ~97 tok/s on an M3 Ultra is a pretty good demonstration of that.
FHILY👑15,378 görüntüleme • 1 ay önce
Qwen3.8-27B vs Ornith-1.5-35B on the Bioluminescent Abyssal Temple challenge. Same prompt. Same Grok Build harness. 4-bit MLX. → Qwen: ~24 tok/s · 33 min → Ornith: ~55 tok/s · 16 min Ornith was over 2× faster and finished in roughly half the time. This is a great example of why raw parameter count doesn’t tell the whole story. Qwen vs Ornith is getting seriously interesting.
FHILY👑23,160 görüntüleme • 2 ay önce
This is how I create a motion design workflow fully AI-powered with Claude and Higgsfield MCP. 1. Start by pulling references via Pinterest API in Claude Code 2. Generate a 6-scene storyboard through GPT Images 2.0 3. In Higgsfield MCP, then render it straight to video in Seedance 2.0 All connected in one seamless pipeline.
FHILY👑45,001 görüntüleme • 4 ay önce
Everyone is talking about Claude Design… but I keep going back to Gemini. Gemini 3.1 Pro is seriously impressive when it comes to motion design. I actually want to try Claude Design, but after only two prompts, I was already running out of credits. If you care about anti AI slop, Gemini is still winning. With a single prompt, it can transform a static HTML layout into dynamic, polished motion graphics.
FHILY👑25,019 görüntüleme • 3 ay önce
Motion design with Claude + Higgsfield MCP 🧩 Turn a single text prompt into pro-level motion videos. It’s actually insanely simple: 1. Describe your idea to Claude 2. Ask it to generate a storyboard 3. Pick your favorite concept 4. Let Higgsfield MCP render the final video
FHILY👑25,763 görüntüleme • 4 ay önce
Qwen3.8-27B just got absurdly smaller. Unsloth’s new Dynamic v3.0 GGUFs bring the entire 27B model down to 6.19GB at UD-IQ1_S. But there’s a catch: token accuracy doesn’t tell the whole story. On a 41-task agent test: → UD-IQ1_S: 6.19GB · 21/41 · 241 tok/s → UD-Q2_K_XL: 9.83GB · 27/41 · 168 tok/s → UD-Q3_K_XL: 13.1GB · 33/41 · 129 tok/s → UD-Q4_K_M: 16.5GB · 35/41 · 107 tok/s The 1-bit quant retains plenty of token-level accuracy, but over a long agent run, small errors compound. The sweet spot here looks like UD-Q3_K_XL: 13.1GB, 129 tok/s, and 33/41 tasks completed. And the performance takeaway is wild: The bottleneck wasn’t the 5090. It was the size of the weights. Smaller model file → less memory bandwidth pressure → dramatically higher decode speed. 6.19GB for a 27B model is getting ridiculous.
FHILY👑13,372 görüntüleme • 2 ay önce
Two frontier models. Two very different demos. (GPT-5.6 Sol and Claude Fable 5) GPT-5.6 Sol Ultra was pushed in a mathematical direction and generated a Minecraft-style clone in Lean. Yes, that Lean. Meanwhile, Claude Fable 5 took a game development route and built a Minecraft mod featuring a Dark Souls-inspired enemy with multiple attack patterns, combos, and combat logic. They're solving completely different problems, but both are great examples of how capable today's models have become. The pace of progress is remarkable.
FHILY👑19,336 görüntüleme • 3 ay önce
Hey Claude, redesign my landing page and add some 3D depth to it 🔥 — Claude Opus 4.8 ultracode
FHILY👑15,946 görüntüleme • 4 ay önce
How I create 3D-feel websites with expanding images using Claude Code👇 1. Use Claude to plan scroll-based animations + scene flow 2. Start with full-width images that expand on scroll 3. Add zoom-in + scale effects to create depth illusion 4. Layer images to simulate foreground/background (fake 3D) 5. Use parallax so elements move at different speeds 6. Add smooth transitions + ambient motion for immersion 7. Fine-tune timing (0.6s–1.2s) for cinematic feel Save this if you want next-level websites 🚀
FHILY👑10,268 görüntüleme • 4 ay önce
Daha fazla içerik yok.