Video wird geladen...

Video konnte nicht geladen werden

Beim Laden dieses Videos ist ein Problem aufgetreten. Dies könnte an einem vorübergehenden Netzwerkproblem liegen oder das Video ist möglicherweise nicht verfügbar.

Locked one identity across four set changes and let the type carry its own weight. All of it in a single 15s pass. MiniMax H3 x Picsart Challenge is live — $50,000 prize pool, submissions close Sep 23. Enter: MiniMax (official) Picsart #ad #MiniMaxH3 #Picsart

Iris

18,398 subscribers

256,040 Aufrufe • vor 3 Tagen •via X (Twitter)

Wissenschaft & Technologie #ad #MiniMaxH3

Anya Rossi• Live Now

Private livecam show

5 Kommentare

Profilbild von Zeynep

Zeynepvor 3 Tagen

@MiniMax_AI @Picsart Bilgilendirme için teşekkürler

Profilbild von Mehwish kiran

Mehwish kiranvor 2 Tagen

@MiniMax_AI @Picsart Four looks, one identity—that consistency is impressive

Profilbild von meeeehna 🥂

meeeehna 🥂vor 3 Tagen

@MiniMax_AI @Picsart Its all about thoughts

Profilbild von Riley Quinn

Riley Quinnvor 2 Tagen

@MiniMax_AI @Picsart Impressive identity lock across four distinct environments, executed cleanly in a single pass.

Profilbild von Naaz Khan Habibi

Naaz Khan Habibivor 3 Tagen

@MiniMax_AI @Picsart It’s all about avoiding temperature swings condensation can make it easier for bacteria to get in.

Ähnliche Videos

The BIGGEST ever AI Challenge is LIVE now! A $20,000 prize pool, FREE gens, 2 weeks, 6 categories & MiniMax FAST. Post a 10-45s video with #higgsfieldmax and tag Higgsfield AI 🧩 and MiniMax Design (H3) to enter. We're going ALL IN. Retweet to get the Full Entry Guide. Learn more →

The BIGGEST ever AI Challenge is LIVE now! A $20,000 prize pool, FREE gens, 2 weeks, 6 categories & MiniMax FAST. Post a 10-45s video with #higgsfieldmax and tag Higgsfield AI 🧩 and MiniMax Design (H3) to enter. We're going ALL IN. Retweet to get the Full Entry Guide. Learn more →

Higgsfield AI 🧩

536,469 Aufrufe • vor 1 Jahr

WAN 3.0 is live on Picsart! 🔊🔥 Complete 30-second videos in one take, and a lot more 🤩 🎬 Full scenes, not fragments: give your story a real beginning, middle, and end 🎯 Up to 20 references to keep every detail consistent across the whole take 🔊 Sound included: dialogue and ambience already synchronized 💡 Draft in 480p, then go full quality on the final Perfect for product videos, ad concepts, social clips, and ideas that need more time to unfold! Try it here 👉

WAN 3.0 is live on Picsart! 🔊🔥 Complete 30-second videos in one take, and a lot more 🤩 🎬 Full scenes, not fragments: give your story a real beginning, middle, and end 🎯 Up to 20 references to keep every detail consistent across the whole take 🔊 Sound included: dialogue and ambience already synchronized 💡 Draft in 480p, then go full quality on the final Perfect for product videos, ad concepts, social clips, and ideas that need more time to unfold! Try it here 👉

Picsart

11,658 Aufrufe • vor 24 Tagen

MINIMAX H3 JUST REPLACED THE ENTIRE ANIMATION PIPELINE before: a studio quotes $150,000 for a 25-second anime opening like this. a designer, an animator, an editor. 6 weeks across 4 different tools. after: i wrote the concept once and got 25 finished seconds back, the cuts and the card graphics already in the frame. I made this anime opening on MiniMax Design from one original idea. Both characters look the same in every shot, and black, white and red hold from the first frame to the last. The agent broke the concept into shots on its own and handed back the finished sequence. Motion graphics, VFX, anime openings and music videos, all in one platform with local assets and scalable APIs. Created with MiniMax H3 and its Omni Reference. Hats off to MiniMax Design (H3) 【#MiniMaxH3 on #MiniMaxDesign】

MINIMAX H3 JUST REPLACED THE ENTIRE ANIMATION PIPELINE before: a studio quotes $150,000 for a 25-second anime opening like this. a designer, an animator, an editor. 6 weeks across 4 different tools. after: i wrote the concept once and got 25 finished seconds back, the cuts and the card graphics already in the frame. I made this anime opening on MiniMax Design from one original idea. Both characters look the same in every shot, and black, white and red hold from the first frame to the last. The agent broke the concept into shots on its own and handed back the finished sequence. Motion graphics, VFX, anime openings and music videos, all in one platform with local assets and scalable APIs. Created with MiniMax H3 and its Omni Reference. Hats off to MiniMax Design (H3) 【#MiniMaxH3 on #MiniMaxDesign】

Fokki

80,007 Aufrufe • vor 29 Tagen

MiniMax (official) H3 is live in SGLang Diffusion, with day-0 serving support 🎬 This open model matches Seedance 2.0 at 1/3 the cost, or $0 if you run it locally on 2x 5090 or 1 RTX 6000. With SGLang Diffusion, you can build visual concepts, motion design, e-commerce creatives, precise video edits, animation and stylized visuals, all locally on your own machine. No API bill, no waitlist. H3 takes text, image, video and audio in a single context, and generates 5-15s clips at native 2K, 24fps with native stereo audio. SGLang Diffusion also runs on NVIDIA AI Blackwell and Hopper, and AMD MI355X / MI300X. Let's create something with MiniMax H3 👇

MiniMax (official) H3 is live in SGLang Diffusion, with day-0 serving support 🎬 This open model matches Seedance 2.0 at 1/3 the cost, or $0 if you run it locally on 2x 5090 or 1 RTX 6000. With SGLang Diffusion, you can build visual concepts, motion design, e-commerce creatives, precise video edits, animation and stylized visuals, all locally on your own machine. No API bill, no waitlist. H3 takes text, image, video and audio in a single context, and generates 5-15s clips at native 2K, 24fps with native stereo audio. SGLang Diffusion also runs on NVIDIA AI Blackwell and Hopper, and AMD MI355X / MI300X. Let's create something with MiniMax H3 👇

LMSYS Org

100,189 Aufrufe • vor 1 Monat

MiniMax H3 is native in ComfyUI, day zero with the weights. Model highlights: → Text-to-video, prompt only → Image-to-video → First-and-last-frame - control the opening frame, the closing frame, or both → Reference-to-video - carry a subject, a motion, or a voice through the clip → In-place editing - modify a shot you already have instead of regenerating it The capability MiniMax (official) leads with is multimodal context understanding, and it's what collapses those five tasks into one model. H3 takes images, audio, and video together and resolves them against a prompt that explains how they relate. Download the workflows & learn more in our blog 👇

MiniMax H3 is native in ComfyUI, day zero with the weights. Model highlights: → Text-to-video, prompt only → Image-to-video → First-and-last-frame - control the opening frame, the closing frame, or both → Reference-to-video - carry a subject, a motion, or a voice through the clip → In-place editing - modify a shot you already have instead of regenerating it The capability MiniMax (official) leads with is multimodal context understanding, and it's what collapses those five tasks into one model. H3 takes images, audio, and video together and resolves them against a prompt that explains how they relate. Download the workflows & learn more in our blog 👇

ComfyUI

234,430 Aufrufe • vor 1 Monat

Desert Hypercar. One prompt. Pure cinema. I wanted to see how far Studio AI's MiniMax H3 could push a simple idea. Starting from a single prompt, it generated a cinematic desert chase with smooth camera movement, realistic lighting, and premium automotive-commercial vibes—all in minutes. This feels like a concept trailer you'd expect from a high-end car commercial, not an AI workflow. Would you drive this through the dunes? #ad #StudioAI #MiniMaxH3 #AIVideo #AIFilmmaking #CinematicAI

Desert Hypercar. One prompt. Pure cinema. I wanted to see how far Studio AI's MiniMax H3 could push a simple idea. Starting from a single prompt, it generated a cinematic desert chase with smooth camera movement, realistic lighting, and premium automotive-commercial vibes—all in minutes. This feels like a concept trailer you'd expect from a high-end car commercial, not an AI workflow. Would you drive this through the dunes? #ad #StudioAI #MiniMaxH3 #AIVideo #AIFilmmaking #CinematicAI

Aina Ai | Tools & Updates

27,049 Aufrufe • vor 1 Monat

🚀 Sol-H3: MiniMax (official) H3 Video Generation Faster Than Playback 🤩 Five seconds of world. 1.653 seconds to infer. We’re releasing Sol-H3, our fastest end-to-end MiniMax-H3 inference stack yet. On one 8× NVIDIA B300 Blackwell system, it generates five seconds of 1344×768 video with stereo audio in 1.653 seconds. Across 1×, 4×, and 8× B300, Sol-H3 reaches up to a 15.54× speedup versus Base H3. Compared with 50-step Base H3 Dense on the same 8× B300 system, the four-step Sol-H3 profile delivers: • 5s: 18.250s → 1.653s (11.04×) • 10s: 50.660s → 3.732s (13.57×) • 15s: 99.513s → 6.612s (15.05×) Sol-H3 also scales across GPU counts: • 4× B300: 2.918s / 6.993s / 12.542s for 5s / 10s / 15s (12.11–15.54×) • 1× B300: 13.745s / 37.813s / 52.260s for 5s / 10s / 15s (9.45–14.29×) All figures are medians of three measured runs after one warmup at 1344×768 and 24 FPS with stereo audio. Base H3 uses 50 scheduler points (49 DiT forwards); Sol-H3 uses four DiT forwards, so this is a full-profile comparison—not an attention-only runtime change. Sol-H3 uses Dense attention on 1× B300 and SOL with INT8 QKV / FP8 output transport on 4× / 8×. Timing includes text encoding, DiT denoising, and video/audio VAE decoding; model loading, compilation warmup, and final MP4 encoding are excluded. Sol-H3 brings Sol-Engine × Sol-Attn into one full-stack runtime: • dynamic sparse attention with no retraining • fused norm, RoPE, MLP, and sparse-attention setup • fused INT8 QKV / FP8 output communication across 8 GPUs • parallel, batched VAE decoding • precomputed AdaLN caching Inside the stack: • sparse-attention setup: 1.206 → 0.285 ms (−76.4%) • VAE decode: 7.55 → 0.602 s • ~24 GB memory freed per GPU Any MiniMax-H3 few-step LoRA can plug into the same engine, and the code is deployment-friendly under Apache 2.0. For us, the bigger milestone is crossing from “fast generation” into “faster than playback.” That opens the path toward continuous 24 FPS generation and truly interactive video systems. We’re excited to partner with reactor to release Sol-H3 and make it available as an API day-0. Try it now on Reactor: 🔗 Amazing team effort—full credits in the blog. LoveSy Junsong_Chen yitong li Haopeng Li Haocheng Xi Song Han

Enze Xie

204,371 Aufrufe • vor 10 Tagen

Good news - over the past week, submissions have generated millions of impressions across our global community. The creativity is being seen, shared, and celebrated all over the world - and we want to create even more space for it. So we’re extending the awards by one more week. One more week to enter. One more week to be seen. One more week to compete for a $90,000 prize pool. Want to be considered? Make sure your 4 qualifying posts are in by April 26.

Good news - over the past week, submissions have generated millions of impressions across our global community. The creativity is being seen, shared, and celebrated all over the world - and we want to create even more space for it. So we’re extending the awards by one more week. One more week to enter. One more week to be seen. One more week to compete for a $90,000 prize pool. Want to be considered? Make sure your 4 qualifying posts are in by April 26.

OpenArt

1,173,168 Aufrufe • vor 5 Monaten

I plugged MiniMax (official) M3 into Claude Code in under 5 minutes and gave it a real task. most models need hand-holding every few steps. M3 ran 147 iterations on its own. what it actually did: → took a broken CUDA kernel → ran it for 24 hours, autonomously → pushed hardware efficiency from 7.6% to 71.3% → shipped a 9.4x speedup I didn't touch it once across that whole window. the rest of the spec sheet: → 1M token context window → native multimodal from day one → fully open weights pricing: → $20/month for 1.7B tokens → $50/month for 5.1B tokens → text, image, speech, and music all share one pool usually with this kind of capability the bill catches up fast. M3 keeps both lines low at the same time. that's the part that surprised me. and if this is what M3 does at launch, the full rollout is going to be one to watch. what would you build with 1M context? #MiniMax #MiniMaxM3 #ClaudeCode #LLM

I plugged MiniMax (official) M3 into Claude Code in under 5 minutes and gave it a real task. most models need hand-holding every few steps. M3 ran 147 iterations on its own. what it actually did: → took a broken CUDA kernel → ran it for 24 hours, autonomously → pushed hardware efficiency from 7.6% to 71.3% → shipped a 9.4x speedup I didn't touch it once across that whole window. the rest of the spec sheet: → 1M token context window → native multimodal from day one → fully open weights pricing: → $20/month for 1.7B tokens → $50/month for 5.1B tokens → text, image, speech, and music all share one pool usually with this kind of capability the bill catches up fast. M3 keeps both lines low at the same time. that's the part that surprised me. and if this is what M3 does at launch, the full rollout is going to be one to watch. what would you build with 1M context? #MiniMax #MiniMaxM3 #ClaudeCode #LLM

Nelly;

17,576 Aufrufe • vor 3 Monaten

I used to think 3D was way too much work for a quick creative idea Then I tried this workflow in MiniMax Design GPT-6 Astra can work with Blender to build the 3D scene from a natural-language prompt, which honestly changes how I think about 3D creation I don’t need to start by figuring out every Blender operation myself I can just describe the scene I have in mind and let Astra handle the technical side Then I can take that 3D result straight into H3 Max Turbo and turn it into a video And the speed here is no joke 15s video → 20s generation Starting at $0.018/s That makes experimenting with different ideas feel so much less expensive in terms of time But the more I use MiniMax Design, the more I realize 3D is only one part of it GPT Image 2.5 opens up another route — generate the visuals and storyboards first, then use H3 Max Turbo to bring those shots together into a video So now I’m basically playing with two different creative paths: Astra + Blender → 3D → H3 Max Turbo GPT Image 2.5 → Storyboard → H3 Max Turbo And having all of these connected inside MiniMax Design makes the whole process feel much more flexible Honestly, I started this just wanting to see what Astra could do with 3D Now I’m way more curious about what happens when I start mixing all these workflows together

I used to think 3D was way too much work for a quick creative idea Then I tried this workflow in MiniMax Design GPT-6 Astra can work with Blender to build the 3D scene from a natural-language prompt, which honestly changes how I think about 3D creation I don’t need to start by figuring out every Blender operation myself I can just describe the scene I have in mind and let Astra handle the technical side Then I can take that 3D result straight into H3 Max Turbo and turn it into a video And the speed here is no joke 15s video → 20s generation Starting at $0.018/s That makes experimenting with different ideas feel so much less expensive in terms of time But the more I use MiniMax Design, the more I realize 3D is only one part of it GPT Image 2.5 opens up another route — generate the visuals and storyboards first, then use H3 Max Turbo to bring those shots together into a video So now I’m basically playing with two different creative paths: Astra + Blender → 3D → H3 Max Turbo GPT Image 2.5 → Storyboard → H3 Max Turbo And having all of these connected inside MiniMax Design makes the whole process feel much more flexible Honestly, I started this just wanting to see what Astra could do with 3D Now I’m way more curious about what happens when I start mixing all these workflows together

haytam

83,103 Aufrufe • vor 6 Tagen

THIS IS F**KING GOLD this 23-year-old made $30,000 this month and barely opens video editing software. she didn’t get better at keyframing. she stopped doing the manual labor. here’s what she actually built: ChatGPT as the brain, Picsart Flow as the animator. ChatGPT studies what’s working across viral kids’ content and turns those patterns into detailed visual prompts for her own character, feeds them into Picsart Flow running Kling 3.0, sets the render parameters, outputs full Pixar-style 3D clips. all from her browser, under 5 minutes per clip. 30 minutes a day, total. she’s not animating frame by frame. she’s orchestrating a pipeline. kids’ channels pull hundreds of millions of views on short-form content without her building a single 3D model. the cross-platform ad revenue math is what should actually get your attention: $2,500–$7,000 per million views on YouTube under 10 minutes render time per clip 2 to 3 videos posted daily across channels $15,000/month in passive revenue she’s not selling animation skill anymore. she’s selling volume. the character concept is hers. everything from prompt structure to final render is the AI’s job. most creators are still trying to figure out 3D modeling tools. she skipped that entirely and automated the production. full breakdown in the article below Bookmark and follow for more.

THIS IS F**KING GOLD this 23-year-old made $30,000 this month and barely opens video editing software. she didn’t get better at keyframing. she stopped doing the manual labor. here’s what she actually built: ChatGPT as the brain, Picsart Flow as the animator. ChatGPT studies what’s working across viral kids’ content and turns those patterns into detailed visual prompts for her own character, feeds them into Picsart Flow running Kling 3.0, sets the render parameters, outputs full Pixar-style 3D clips. all from her browser, under 5 minutes per clip. 30 minutes a day, total. she’s not animating frame by frame. she’s orchestrating a pipeline. kids’ channels pull hundreds of millions of views on short-form content without her building a single 3D model. the cross-platform ad revenue math is what should actually get your attention: $2,500–$7,000 per million views on YouTube under 10 minutes render time per clip 2 to 3 videos posted daily across channels $15,000/month in passive revenue she’s not selling animation skill anymore. she’s selling volume. the character concept is hers. everything from prompt structure to final render is the AI’s job. most creators are still trying to figure out 3D modeling tools. she skipped that entirely and automated the production. full breakdown in the article below Bookmark and follow for more.

Mayor.

74,977 Aufrufe • vor 10 Tagen

She's selling a $20,000 a month plan from a pool chair, one AirPod in. She says all you need is a phone. She's tapping a full iPad in landscape, silver nails flashing. Search cartoon kids shows. Point at Doggyland, 6.6 million views. Copy the ABC song transcript from the description. Paste it into ChatGPT. "Make me a prompt for a show just like this." Copy what it gives back. Open Picsart Flow. Paste. Pause at 0:38. The model in the dropdown is "Sora 2." Sora 2 isn't in Picsart. Sora 2 isn't anywhere. The resolution under it reads 1280x720. The clip is 12 seconds. Two seconds later the screen fills with bouncing letters, an elephant, a parrot, and a fully mixed song with the lyrics synced underneath. "And this is all in 4K." The prompt only asked for animation. It never asked for a song. A 12 second 720p clip became a finished music video between two taps. That didn't happen on screen. None of this is the point. The point is the workflow is a prop. There is no Sora 2 button that prints a 4K music video. The real thing is four cheap parts. Claude breaks the transcript into scenes and writes the visual prompt for each. Suno turns the lyrics into the song. Kling or Luma renders the clips. ffmpeg stitches them and Whisper syncs the captions. Pennies per call. One evening of Python. She wants you to comment FLOW so she can dm you a template that doesn't run. I rebuilt the real factory, Claude to Suno to Kling. Comment 720 and I'll dm it free.

She's selling a $20,000 a month plan from a pool chair, one AirPod in. She says all you need is a phone. She's tapping a full iPad in landscape, silver nails flashing. Search cartoon kids shows. Point at Doggyland, 6.6 million views. Copy the ABC song transcript from the description. Paste it into ChatGPT. "Make me a prompt for a show just like this." Copy what it gives back. Open Picsart Flow. Paste. Pause at 0:38. The model in the dropdown is "Sora 2." Sora 2 isn't in Picsart. Sora 2 isn't anywhere. The resolution under it reads 1280x720. The clip is 12 seconds. Two seconds later the screen fills with bouncing letters, an elephant, a parrot, and a fully mixed song with the lyrics synced underneath. "And this is all in 4K." The prompt only asked for animation. It never asked for a song. A 12 second 720p clip became a finished music video between two taps. That didn't happen on screen. None of this is the point. The point is the workflow is a prop. There is no Sora 2 button that prints a 4K music video. The real thing is four cheap parts. Claude breaks the transcript into scenes and writes the visual prompt for each. Suno turns the lyrics into the song. Kling or Luma renders the clips. ffmpeg stitches them and Whisper syncs the captions. Pennies per call. One evening of Python. She wants you to comment FLOW so she can dm you a template that doesn't run. I rebuilt the real factory, Claude to Suno to Kling. Comment 720 and I'll dm it free.

Aeron

254,852 Aufrufe • vor 2 Monaten

He stopped job hunting and now pulls $12,000 a month from AI cartoons Stop looking for a job and do this instead. Every viral kids video you have seen lately is already made by AI. And YouTube pays anywhere from $2,000 to $12,000 for a single million views. Here is how he does it in 30 seconds. Ask ChatGPT for a 1 minute Pixar style kids cartoon prompt. Copy the whole thing. Open Picsart Flow and start a new workflow. Paste it in, let the copilot run on auto and just sit back. Out comes a clean 4K cartoon clip that looks straight out of a studio. You never film anything or show your face once. Post these daily and the views turn into real money. The quality is insane and barely anyone is doing it yet. Save this before that changes.

He stopped job hunting and now pulls $12,000 a month from AI cartoons Stop looking for a job and do this instead. Every viral kids video you have seen lately is already made by AI. And YouTube pays anywhere from $2,000 to $12,000 for a single million views. Here is how he does it in 30 seconds. Ask ChatGPT for a 1 minute Pixar style kids cartoon prompt. Copy the whole thing. Open Picsart Flow and start a new workflow. Paste it in, let the copilot run on auto and just sit back. Out comes a clean 4K cartoon clip that looks straight out of a studio. You never film anything or show your face once. Post these daily and the views turn into real money. The quality is insane and barely anyone is doing it yet. Save this before that changes.

winkle.

120,093 Aufrufe • vor 3 Monaten

Qwen3.8 27B wrote this game, Minimax H3 turned my gameplay into this video, and Minimax Music 3 wrote the soundtrack. Three local models, one apartment, zero cloud. Sound on. The chain: Qwen3.8 built the shooter in DeepSeek Harness, thats the 687 step log from my last post. I recorded a run, gave the recording to H3 as a video reference, and it rebuilt the level in its own render. Then Music 3 wrote the EBM track under it, two minutes per candidate, I picked one out of a stack. I wanted a frame for frame redraw first and it did not happen. The reference mode treats the anchor as a hint, so stitched chunks tore at the seams. The cut that worked: 10 seconds rebuilt straight from my recording, then 25 seconds where the model continues the level on its own, every chunk starting hard from the previous last frame. A bit visible seams, and honestly the invented half moves better than my footage. 35 seconds at 1344x768 took about an hour on the rig. The reference part burns roughly 7x the compute per second of plain generation, the free continuation is the cheap half. The game engine ate two 3090s. The video and the music took the other cards. One rig in one apartment now covers the whole credits list: code, footage, score. 18 months ago each of these steps was a company. Now they hand work to each other on my desk overnight. Qwen MiniMax (official)

Qwen3.8 27B wrote this game, Minimax H3 turned my gameplay into this video, and Minimax Music 3 wrote the soundtrack. Three local models, one apartment, zero cloud. Sound on. The chain: Qwen3.8 built the shooter in DeepSeek Harness, thats the 687 step log from my last post. I recorded a run, gave the recording to H3 as a video reference, and it rebuilt the level in its own render. Then Music 3 wrote the EBM track under it, two minutes per candidate, I picked one out of a stack. I wanted a frame for frame redraw first and it did not happen. The reference mode treats the anchor as a hint, so stitched chunks tore at the seams. The cut that worked: 10 seconds rebuilt straight from my recording, then 25 seconds where the model continues the level on its own, every chunk starting hard from the previous last frame. A bit visible seams, and honestly the invented half moves better than my footage. 35 seconds at 1344x768 took about an hour on the rig. The reference part burns roughly 7x the compute per second of plain generation, the free continuation is the cheap half. The game engine ate two 3090s. The video and the music took the other cards. One rig in one apartment now covers the whole credits list: code, footage, score. 18 months ago each of these steps was a company. Now they hand work to each other on my desk overnight. Qwen MiniMax (official)

Alexey Fateev

34,605 Aufrufe • vor 1 Monat

hy3 vs mimo-v2.5 vs deepseek v4 flash vs minimax m3 the four models on top of the openrouter leaderboard by tokens this week: #1 hy3 (Tencent Hy) – 7.5t #2 mimo-v2.5 (Xiaomi MiMo) – 6.56t #3 deepseek v4 flash (DeepSeek) – 5.24t #4 minimax m3 (MiniMax (official)) – 4.21t so we tested them. 3 prompts, single-file html, Three.js from a cdn, fully procedural, no external assets. all run via AI/ML API each prompt is a transparent cutaway machine that has to be mechanically correct, not decorative: • 4-stroke engine with full oil circulation – slider-crank kinematics, cam at 2:1, valve lift driven by lobes, oil loop from sump to gallery to big-end • watt walking-beam steam engine – four-bar vector-loop closure, eccentric-driven slide valve, steam events synced to real port position • francis reaction water turbine – 20 guide vanes on a regulating ring, 17 lofted runner blades, gpu particle advection, precessing vortex rope at part load the takeaway up front: none of the four cleared all three scenes on the first attempt. but the price spread between them is roughly 70x – hy3 fixed included costs less than two cents overall results (summed across all 3 scenes): cost #1 hy3 – $0.016 #2 deepseek v4 flash – $0.025 #3 mimo-v2.5 – $0.97 #4 minimax m3 – $1.17 tokens #1 hy3 – 19,326 #2 deepseek v4 flash – 63,126 #3 mimo-v2.5 – 322,523 #4 minimax m3 – 702,900 lines of code #1 hy3 – 1,047 #2 mimo-v2.5 – 2,759 #3 deepseek v4 flash – 3,273 #4 minimax m3 – 3,354 scenes needing a second attempt #1 hy3 – 1 (engine) #1 mimo-v2.5 – 1 (turbine) #1 minimax m3 – 1 (turbine) #4 deepseek v4 flash – 2 (steam engine, turbine) observations: 1. the token spread is the real story – minimax burns 36x hy3's tokens and lands in the same place, one retry, ~3.3k lines 2. hy3 is the outlier on density: 1,047 lines total, fewest tokens, cheapest run, and only one scene needed a second pass. deepseek is the opposite trade – near-hy3 pricing but the most retries 3. mimo and minimax seem to overthink instead of writing the code. minimax spent 359.1k tokens on the steam engine and produced 1,346 lines – the tokens are going somewhere other than the file 4. the francis turbine broke three of the four. the spec that separates them is the one with 20 linked guide vanes and gpu particle advection, not the one with the most parts overall impression: none of these models excelled at any of the tasks we gave them. but they were close, and they were extremely cheap. the gap that matters isn't quality anymore – it's that hy3 ran all three scenes for less than two cents while the frontier labs charge dollars for the same work right now you pick these because they're good for the zero price you pay. soon that's something openai and anthropic will have to think about follow thehype. for 24/7 ai news, analysis and breakdowns

hy3 vs mimo-v2.5 vs deepseek v4 flash vs minimax m3 the four models on top of the openrouter leaderboard by tokens this week: #1 hy3 (Tencent Hy) – 7.5t #2 mimo-v2.5 (Xiaomi MiMo) – 6.56t #3 deepseek v4 flash (DeepSeek) – 5.24t #4 minimax m3 (MiniMax (official)) – 4.21t so we tested them. 3 prompts, single-file html, Three.js from a cdn, fully procedural, no external assets. all run via AI/ML API each prompt is a transparent cutaway machine that has to be mechanically correct, not decorative: • 4-stroke engine with full oil circulation – slider-crank kinematics, cam at 2:1, valve lift driven by lobes, oil loop from sump to gallery to big-end • watt walking-beam steam engine – four-bar vector-loop closure, eccentric-driven slide valve, steam events synced to real port position • francis reaction water turbine – 20 guide vanes on a regulating ring, 17 lofted runner blades, gpu particle advection, precessing vortex rope at part load the takeaway up front: none of the four cleared all three scenes on the first attempt. but the price spread between them is roughly 70x – hy3 fixed included costs less than two cents overall results (summed across all 3 scenes): cost #1 hy3 – $0.016 #2 deepseek v4 flash – $0.025 #3 mimo-v2.5 – $0.97 #4 minimax m3 – $1.17 tokens #1 hy3 – 19,326 #2 deepseek v4 flash – 63,126 #3 mimo-v2.5 – 322,523 #4 minimax m3 – 702,900 lines of code #1 hy3 – 1,047 #2 mimo-v2.5 – 2,759 #3 deepseek v4 flash – 3,273 #4 minimax m3 – 3,354 scenes needing a second attempt #1 hy3 – 1 (engine) #1 mimo-v2.5 – 1 (turbine) #1 minimax m3 – 1 (turbine) #4 deepseek v4 flash – 2 (steam engine, turbine) observations: 1. the token spread is the real story – minimax burns 36x hy3's tokens and lands in the same place, one retry, ~3.3k lines 2. hy3 is the outlier on density: 1,047 lines total, fewest tokens, cheapest run, and only one scene needed a second pass. deepseek is the opposite trade – near-hy3 pricing but the most retries 3. mimo and minimax seem to overthink instead of writing the code. minimax spent 359.1k tokens on the steam engine and produced 1,346 lines – the tokens are going somewhere other than the file 4. the francis turbine broke three of the four. the spec that separates them is the one with 20 linked guide vanes and gpu particle advection, not the one with the most parts overall impression: none of these models excelled at any of the tasks we gave them. but they were close, and they were extremely cheap. the gap that matters isn't quality anymore – it's that hy3 ran all three scenes for less than two cents while the frontier labs charge dollars for the same work right now you pick these because they're good for the zero price you pay. soon that's something openai and anthropic will have to think about follow thehype. for 24/7 ai news, analysis and breakdowns

thehype.

17,145 Aufrufe • vor 2 Monaten

A Nobel Prize-winning economist was seated in front of a television set from 1977 and was asked to argue for the idea that anyone should be allowed to own anything. Almost no one watches it. Friedrich Hayek is appearing on Firing Line three years after winning the Nobel Prize to defend the idea of private property against a live panel. There's no editing and no commercial break just four men in suits, one camera, and one hour of argument recorded on tape. Look at the stage at which he goes from property to knowledge. The argument is not that owners are entitled to it, but that no one can have sufficient information without it. People debate property online, and he did so on live television.

A Nobel Prize-winning economist was seated in front of a television set from 1977 and was asked to argue for the idea that anyone should be allowed to own anything. Almost no one watches it. Friedrich Hayek is appearing on Firing Line three years after winning the Nobel Prize to defend the idea of private property against a live panel. There's no editing and no commercial break just four men in suits, one camera, and one hour of argument recorded on tape. Look at the stage at which he goes from property to knowledge. The argument is not that owners are entitled to it, but that no one can have sufficient information without it. People debate property online, and he did so on live television.

the lich

344,208 Aufrufe • vor 1 Monat

🚨 ASI:One is officially live across London. And so is your chance to win a luxury 5-star two-night stay in London with afternoon tea. Do this to enter the competiton: (1) Look out for ASI:One buses + billboards. (2) Snap a selfie when you’ll spot our adverts (your face + our ad needs to be visible). (3) Post it on X using and tag Fetch.ai in the caption. Boost your chance of winning: Travel across the city and snap more photos of ASI:One ads. We’ll keep track of all the selfies + tags and will reveal the winner on the 1st of December. Challenge runs till 30/11/2025. Good luck!

🚨 ASI:One is officially live across London. And so is your chance to win a luxury 5-star two-night stay in London with afternoon tea. Do this to enter the competiton: (1) Look out for ASI:One buses + billboards. (2) Snap a selfie when you’ll spot our adverts (your face + our ad needs to be visible). (3) Post it on X using and tag Fetch.ai in the caption. Boost your chance of winning: Travel across the city and snap more photos of ASI:One ads. We’ll keep track of all the selfies + tags and will reveal the winner on the 1st of December. Challenge runs till 30/11/2025. Good luck!

Fetch.ai

13,225 Aufrufe • vor 10 Monaten

SITUATION EXPLAINED: MiniMax released H3 with open weights, and it's already #1 in video editing. • Priced at $7.80 per minute of 2K video, against Seedance 2.0 at $22.45 and Kling 3.0 at $20.16 • It ranks #1 on Artificial Analysis for video editing, #2 for text-to-video behind Gemini Omni Flash, and #3 for image-to-video • H3 takes text, images, video, and audio in a single context and generates 5-15 second clips at native 2K, 24fps, with native stereo audio • MiniMax is releasing the weights under its Community License within days, which would make it the strongest open-weights video model by a wide margin • It launched the same week ByteDance rolled out Seedance 2.5, which now generates 30-second 4K video from a single prompt • Seedance 2.5's face consistency across a full 30 seconds is the visible jump, earlier models drifted into a different face partway through • Voices are still the obvious tell, the video reads as real long before the audio does • Smaller labs building purpose-built video models are now beating frontier labs at this specific task Theo Jaffee: "This is gonna be one of the biggest examples of things that people are gonna do with these models, historical simulations, basically. Nostalgia videos for zoomers."

SITUATION EXPLAINED: MiniMax released H3 with open weights, and it's already #1 in video editing. • Priced at $7.80 per minute of 2K video, against Seedance 2.0 at $22.45 and Kling 3.0 at $20.16 • It ranks #1 on Artificial Analysis for video editing, #2 for text-to-video behind Gemini Omni Flash, and #3 for image-to-video • H3 takes text, images, video, and audio in a single context and generates 5-15 second clips at native 2K, 24fps, with native stereo audio • MiniMax is releasing the weights under its Community License within days, which would make it the strongest open-weights video model by a wide margin • It launched the same week ByteDance rolled out Seedance 2.5, which now generates 30-second 4K video from a single prompt • Seedance 2.5's face consistency across a full 30 seconds is the visible jump, earlier models drifted into a different face partway through • Voices are still the obvious tell, the video reads as real long before the audio does • Smaller labs building purpose-built video models are now beating frontier labs at this specific task Theo Jaffee: "This is gonna be one of the biggest examples of things that people are gonna do with these models, historical simulations, basically. Nostalgia videos for zoomers."

MTS

55,187 Aufrufe • vor 1 Monat

ComfyUI is now the first and only commercial license distributor for MiniMax (official)'s generative media models. MiniMax open weights have been downloaded through Comfy close to 20 million times. Not one of those downloads paid the lab a cent. That is the open-weight problem. Training frontier models is expensive. Adoption is not revenue. If open models are going to keep pace with the biggest closed labs, the labs need a way to get paid. So we built the bridge. Enterprises, studios, and agencies license through us. The labs get incentive to keep researching. Independent creators keep building exactly as they do now, funded by the organizations with budgets. A license only gets you the model. Our Forward Deployed Creatives embed with your team, train LoRAs on your style, build production workflows in your environment, and leave your people able to run and extend all of it. What they build is yours to keep. MiniMax is where we start, not where we stop. My assertion is that open-weight labs that survive the next two years will not be the ones with the best benchmarks. They will be the ones that solve distribution to paying customers first. Try for Free | Request a License today ⬇️

ComfyUI is now the first and only commercial license distributor for MiniMax (official)'s generative media models. MiniMax open weights have been downloaded through Comfy close to 20 million times. Not one of those downloads paid the lab a cent. That is the open-weight problem. Training frontier models is expensive. Adoption is not revenue. If open models are going to keep pace with the biggest closed labs, the labs need a way to get paid. So we built the bridge. Enterprises, studios, and agencies license through us. The labs get incentive to keep researching. Independent creators keep building exactly as they do now, funded by the organizations with budgets. A license only gets you the model. Our Forward Deployed Creatives embed with your team, train LoRAs on your style, build production workflows in your environment, and leave your people able to run and extend all of it. What they build is yours to keep. MiniMax is where we start, not where we stop. My assertion is that open-weight labs that survive the next two years will not be the ones with the best benchmarks. They will be the ones that solve distribution to paying customers first. Try for Free | Request a License today ⬇️

Yoland Yan

14,896 Aufrufe • vor 21 Tagen

ResNet by hand ✍️ ~ 10 steps walkthrough below "Deep Residual Learning for Image Recognition" (Kaiming He, CVPR 2016) is among the most cited papers in all of deep learning. Why does it matter so much? It fixed the exploding and vanishing gradients that kept deep networks from being deep, and made thousands of layers possible. How simple was the fix? An identity matrix. Goal: push three input vectors through a residual block, then through a transformer encoder block, filling in every cell yourself. = 1. Given = A mini batch of three input vectors, 3D, and the weights of the layers ahead. = 2. Linear layer = Let us multiply by the weights, add the bias, and apply ReLU so negatives become 0. Three feature vectors out. This is F(X). = 3. Concatenate = Now the trick. Stack an identity matrix beside the second layer's weights, and stack the input vectors under the features. Draw the lines between rows and columns: those are the skip connections. The identity is the residual. = 4. Linear layer + identity = We multiply the two stacked matrices. The identity carries X straight through while the weights transform it, so a single multiplication computes F(X) + X. Apply ReLU and hand it to the next block. Now watch the same trick inside a transformer, first in attention. = 5. Attention = Let us take three input vectors in 2D, compute the attention matrix, and multiply to get attention weighted vectors. = 6. Concatenate = We stack two identities this time, two residuals, which is how you get 1 + 1, and stack the input vectors with the attention weighted ones. = 7. Add = Multiply the stacked matrices. The identity adds attention to its own input, across the columns, which is how positions get combined. And again in the feed forward layer. = 8. First layer = Let us multiply by the feed forward weights and bias, then ReLU. Three feature vectors. = 9. Concatenate = Stack and link exactly as in step 3: the residual again. = 10. Second layer + identity = We multiply, apply ReLU, and pass the result to the next encoder block. This identity adds across the rows, combining features rather than positions. Takeaway: one simple "add" is what made really deep networks possible. 💾 Save this post!

ResNet by hand ✍️ ~ 10 steps walkthrough below "Deep Residual Learning for Image Recognition" (Kaiming He, CVPR 2016) is among the most cited papers in all of deep learning. Why does it matter so much? It fixed the exploding and vanishing gradients that kept deep networks from being deep, and made thousands of layers possible. How simple was the fix? An identity matrix. Goal: push three input vectors through a residual block, then through a transformer encoder block, filling in every cell yourself. = 1. Given = A mini batch of three input vectors, 3D, and the weights of the layers ahead. = 2. Linear layer = Let us multiply by the weights, add the bias, and apply ReLU so negatives become 0. Three feature vectors out. This is F(X). = 3. Concatenate = Now the trick. Stack an identity matrix beside the second layer's weights, and stack the input vectors under the features. Draw the lines between rows and columns: those are the skip connections. The identity is the residual. = 4. Linear layer + identity = We multiply the two stacked matrices. The identity carries X straight through while the weights transform it, so a single multiplication computes F(X) + X. Apply ReLU and hand it to the next block. Now watch the same trick inside a transformer, first in attention. = 5. Attention = Let us take three input vectors in 2D, compute the attention matrix, and multiply to get attention weighted vectors. = 6. Concatenate = We stack two identities this time, two residuals, which is how you get 1 + 1, and stack the input vectors with the attention weighted ones. = 7. Add = Multiply the stacked matrices. The identity adds attention to its own input, across the columns, which is how positions get combined. And again in the feed forward layer. = 8. First layer = Let us multiply by the feed forward weights and bias, then ReLU. Three feature vectors. = 9. Concatenate = Stack and link exactly as in step 3: the residual again. = 10. Second layer + identity = We multiply, apply ReLU, and pass the result to the next encoder block. This identity adds across the rows, combining features rather than positions. Takeaway: one simple "add" is what made really deep networks possible. 💾 Save this post!

Tom Yeh

18,152 Aufrufe • vor 1 Monat