🚨 Qwen-Image 2.1
Alibaba just dropped Qwen-Image 2.1 — a unified text-to-image + image editing model with only 7B parameters.
🔥 Native 2K
🔥 Up to 10 reference images
🔥 Better identity & consistency
🔥 Masks/scribbles for local editing
🔥 Native RGBA + transparency
🔥 ComfyUI support from day one
The really interesting part: RGBA editing and object extraction from regular RGB images. Need to test this one.
💾 VRAM-wise, 12 GB looks like a realistic target with quantization/offloading. 16 GB should be very comfortable.
The previous Qwen Image was ~20B. This one is just 7B.
🔗 https://modelscope.ai/models/Qwen/Qwen-Image-2.1/summary
Alibaba just dropped Qwen-Image 2.1 — a unified text-to-image + image editing model with only 7B parameters.
🔥 Native 2K
🔥 Up to 10 reference images
🔥 Better identity & consistency
🔥 Masks/scribbles for local editing
🔥 Native RGBA + transparency
🔥 ComfyUI support from day one
The really interesting part: RGBA editing and object extraction from regular RGB images. Need to test this one.
💾 VRAM-wise, 12 GB looks like a realistic target with quantization/offloading. 16 GB should be very comfortable.
The previous Qwen Image was ~20B. This one is just 7B.
🔗 https://modelscope.ai/models/Qwen/Qwen-Image-2.1/summary
🔥569👍565❤521🎉516
Opus 5.5 vs GPT-6 Sol & Luna — same day!
Anthropic dropped Opus 5.5, while OpenAI released GPT-6 Sol and Luna.
Opus 5.5: $4 / $20, AA Index 58 (#1), Terminal-Bench 66.4%, GDPval-AA 1846 Elo. Anthropic also promises less “Claude-style” fluff.
Sol: $2 / $10, 1.05M context, AA 48. DeepSWE: 68.8% at $2.74/task.
Luna: $0.10 / $0.50, AA 37. DeepSWE: 66.6% at just $0.22/task — almost the same score as Sol for 12x less.
The catch with Opus: tokens are 20% cheaper, but usage is much higher — 119K tokens/task vs 73K for Opus 5 and 31K for Sol. CodeRabbit saw 40–60% more tokens on code reviews.
On AA, Opus 5.5 medium vs Sol max: coding 52.5% vs 43.9%, documents +159 Elo for Opus, business processes roughly tied. Cost: $1.34 vs $1.06/task.
Bottom line: Opus 5.5 is currently the smartest model and 2.5x cheaper than Fable. Luna is the big story for high-volume classification and extraction: 10x cheaper than Haiku. Sol is the mid-range option for mass workloads. Sonnet 5.5 reportedly arrives in a couple of weeks.
Anthropic dropped Opus 5.5, while OpenAI released GPT-6 Sol and Luna.
Opus 5.5: $4 / $20, AA Index 58 (#1), Terminal-Bench 66.4%, GDPval-AA 1846 Elo. Anthropic also promises less “Claude-style” fluff.
Sol: $2 / $10, 1.05M context, AA 48. DeepSWE: 68.8% at $2.74/task.
Luna: $0.10 / $0.50, AA 37. DeepSWE: 66.6% at just $0.22/task — almost the same score as Sol for 12x less.
The catch with Opus: tokens are 20% cheaper, but usage is much higher — 119K tokens/task vs 73K for Opus 5 and 31K for Sol. CodeRabbit saw 40–60% more tokens on code reviews.
On AA, Opus 5.5 medium vs Sol max: coding 52.5% vs 43.9%, documents +159 Elo for Opus, business processes roughly tied. Cost: $1.34 vs $1.06/task.
Bottom line: Opus 5.5 is currently the smartest model and 2.5x cheaper than Fable. Luna is the big story for high-volume classification and extraction: 10x cheaper than Haiku. Sol is the mid-range option for mass workloads. Sonnet 5.5 reportedly arrives in a couple of weeks.
🔥188👍160❤159🎉150
Media is too big
VIEW IN TELEGRAM
Runway has released a plugin for DaVinci Resolve, Adobe Premiere, and After Effects.
It lets you generate, extend, and upscale videos and images directly on the timeline. Video editing, including HDR upgrades, is also available and is powered by Aleph 2.
Available video models: Runway Gen-4.5, Veo 3.1, Seedance 2.5, and Kling 3.0 Pro.
For images: Gen-4 Image, Nano Banana, and GPT Image 2.
The plugin is available on Windows and macOS. For Linux, ask Claude.
It lets you generate, extend, and upscale videos and images directly on the timeline. Video editing, including HDR upgrades, is also available and is powered by Aleph 2.
Available video models: Runway Gen-4.5, Veo 3.1, Seedance 2.5, and Kling 3.0 Pro.
For images: Gen-4 Image, Nano Banana, and GPT Image 2.
The plugin is available on Windows and macOS. For Linux, ask Claude.
🎉191🔥185👍171❤165
Media is too big
VIEW IN TELEGRAM
Kling 4.0 launches in October, while Kling 4.0 Flash is already available to Ultra Yearly subscribers.
Kling 4.0 brings:
up to 30 sec per generation;
up to 10 keyframes;
up to 15 multimodal references;
improved Omni Reference and video editing;
up to 4K;
10-bit HDR;
better stereo audio and lip sync;
video extension;
more languages, accents, and dialects.
It can edit characters, facial expressions, movement, objects, backgrounds, visual style, camera angles, and camera motion.
Edit mode supports one main video plus up to four video references.
Kling 4.0 Flash is a lighter, faster version already in early access.
There’s still no official 4.0 vs Flash comparison, but preliminary info suggests:
Kling 4.0
up to 30 sec;
up to 4K;
15 references;
10 keyframes;
maximum quality and control.
Kling 4.0 Flash
reportedly up to 20 sec;
reportedly 720p;
some advanced features may be limited.
Kling 4.0 brings:
up to 30 sec per generation;
up to 10 keyframes;
up to 15 multimodal references;
improved Omni Reference and video editing;
up to 4K;
10-bit HDR;
better stereo audio and lip sync;
video extension;
more languages, accents, and dialects.
It can edit characters, facial expressions, movement, objects, backgrounds, visual style, camera angles, and camera motion.
Edit mode supports one main video plus up to four video references.
Kling 4.0 Flash is a lighter, faster version already in early access.
There’s still no official 4.0 vs Flash comparison, but preliminary info suggests:
Kling 4.0
up to 30 sec;
up to 4K;
15 references;
10 keyframes;
maximum quality and control.
Kling 4.0 Flash
reportedly up to 20 sec;
reportedly 720p;
some advanced features may be limited.
❤135🔥119🎉119👍109
Searching for a Suno alternative?
This Vortex walkthrough explores Mozart's music workflow through one original electronic track, from Hans V2 and Vibe to Layers, Studio and Personas. The focus is creative control and what happens after generation, without declaring a winner on sound quality.
https://youtu.be/MW5eGyZ1Xbg
This Vortex walkthrough explores Mozart's music workflow through one original electronic track, from Hans V2 and Vibe to Layers, Studio and Personas. The focus is creative control and what happens after generation, without declaring a winner on sound quality.
https://youtu.be/MW5eGyZ1Xbg
YouTube
The best alternative to SUNO | Mozart
Try Mozart AI here https://mozartai.com/?aff=vortex
Searching for a Suno alternative? This Vortex walkthrough explores Mozart's music workflow through one original electronic track, from Hans V2 and Vibe to Layers, Studio and Personas. The focus is creative…
Searching for a Suno alternative? This Vortex walkthrough explores Mozart's music workflow through one original electronic track, from Hans V2 and Vibe to Layers, Studio and Personas. The focus is creative…
👍248🎉232🔥223❤214
The comeback nobody expected.
Gemini 4 just dropped, and it’s absolutely cooking the competition on benchmarks.
Even I didn’t expect Google to come back THIS hard.
Forget the dragon memes. The real dragon just entered the arena.
Gemini 4 Argon is here.
Google’s new frontier model is built for complex coding, enterprise knowledge work, and cybersecurity workflows.
And this part is wild: up to a 1M token output limit.
Argon is rolling out first to trusted testers through Google’s Fairwind Program, with broader access for developers, enterprises, and consumers coming later.
Google is seriously back.
Gemini 4 just dropped, and it’s absolutely cooking the competition on benchmarks.
Even I didn’t expect Google to come back THIS hard.
Forget the dragon memes. The real dragon just entered the arena.
Gemini 4 Argon is here.
Google’s new frontier model is built for complex coding, enterprise knowledge work, and cybersecurity workflows.
And this part is wild: up to a 1M token output limit.
Argon is rolling out first to trusted testers through Google’s Fairwind Program, with broader access for developers, enterprises, and consumers coming later.
Google is seriously back.
🔥374🎉371❤368👍365
Black Forest Labs just dropped FLUX 3 Image.
Generation and editing are now combined in one model.
What’s new:
Bounding boxes for precise placement and local edits: move, delete, replace, or modify specific objects, faces, text, and elements.
Pixel-perfect editing aims to preserve the rest of the image, though shadows, reflections, and nearby lighting may still shift.
Native 4K rendering up to ~16 MP, versus ~4 MP in FLUX.2.
Up to 10 references, plus one endpoint for both generation and editing.
Grounding is enabled by default, so the model can use web/image search for current context.
Pricing:
768 × 768 — $0.041
1K — $0.048
2K — $0.100
4K — $0.607
There’s also a 1K promo price of $0.024.
Commercial Weights are already available for self-hosting and fine-tuning.
Open Weights are coming “in the next few weeks.” Model size, VRAM requirements, quantization options, and final licensing are still unknown.
Available via BFL Playground/API and OpenArt.
https://bfl.ai/models/flux-3-image
Generation and editing are now combined in one model.
What’s new:
Bounding boxes for precise placement and local edits: move, delete, replace, or modify specific objects, faces, text, and elements.
Pixel-perfect editing aims to preserve the rest of the image, though shadows, reflections, and nearby lighting may still shift.
Native 4K rendering up to ~16 MP, versus ~4 MP in FLUX.2.
Up to 10 references, plus one endpoint for both generation and editing.
Grounding is enabled by default, so the model can use web/image search for current context.
Pricing:
768 × 768 — $0.041
1K — $0.048
2K — $0.100
4K — $0.607
There’s also a 1K promo price of $0.024.
Commercial Weights are already available for self-hosting and fine-tuning.
Open Weights are coming “in the next few weeks.” Model size, VRAM requirements, quantization options, and final licensing are still unknown.
Available via BFL Playground/API and OpenArt.
https://bfl.ai/models/flux-3-image
👍316🔥315❤308🎉285
Gemini 4 Argon... dead on arrival?
Google announced its new frontier model just DAYS ago, and almost nobody can even use it yet. Access is still limited to trusted cyber defenders.
And the hype cycle has already moved on.
GPT-6 Bel rumors are exploding.
Fable 5.5 leaks are everywhere.
Neither Bel nor Fable 5.5 is officially confirmed right now.
Argon didn't even get a full week in the spotlight.
AI model lifespans aren't months anymore.
They're DAYS.
Google announced its new frontier model just DAYS ago, and almost nobody can even use it yet. Access is still limited to trusted cyber defenders.
And the hype cycle has already moved on.
GPT-6 Bel rumors are exploding.
Fable 5.5 leaks are everywhere.
Neither Bel nor Fable 5.5 is officially confirmed right now.
Argon didn't even get a full week in the spotlight.
AI model lifespans aren't months anymore.
They're DAYS.
❤194👍179🔥172🎉169
This media is not supported in your browser
VIEW IN TELEGRAM
LTX just dropped IC-LoRA Layout to Render for LTX-2.5 22B.
~1.31 GB, rank 128, trained to use a 3D viewport/playblast as a control signal.
Neural rendering in action — and open source.
Ready-to-use two-pass ComfyUI workflow:
LTX-2.5_ICLoRA_Layout_To_Render_Two_Stage_Distilled.json
First pass: 960×544, 8 steps.
Second: 1920×1088, 3 steps after 2× latent upscale.
Input: a Blender playblast + first frame in your target style. Best approach: run the first playblast frame through image-to-image first.
Requires distilled LTX-2.5 weights, Gemma 4 12B encoder, video/audio VAE and spatial upscaler.
They also have Alpha Gen, SDR→HDR, Native Resolution, Refine/Restore and Layout to Render — Lightricks is positioning LTX as an open-source post-production/VFX toolkit.
~1.31 GB, rank 128, trained to use a 3D viewport/playblast as a control signal.
Neural rendering in action — and open source.
Ready-to-use two-pass ComfyUI workflow:
LTX-2.5_ICLoRA_Layout_To_Render_Two_Stage_Distilled.json
First pass: 960×544, 8 steps.
Second: 1920×1088, 3 steps after 2× latent upscale.
Input: a Blender playblast + first frame in your target style. Best approach: run the first playblast frame through image-to-image first.
Requires distilled LTX-2.5 weights, Gemma 4 12B encoder, video/audio VAE and spatial upscaler.
They also have Alpha Gen, SDR→HDR, Native Resolution, Refine/Restore and Layout to Render — Lightricks is positioning LTX as an open-source post-production/VFX toolkit.
👍98🎉95🔥94❤93
Google just casually dropped Nano Banana 2.1 in Google Flow 🍌
It’s fast, prompt following looks better, and text + capitalization are noticeably cleaner than Nano Banana 2 and Pro.
Some testers say it’s around 2x faster than GPT Image 2.5.
Overall, Nano Banana 2.1 feels like it’s all about speed — but I still think GPT Images delivers better quality.
It’s fast, prompt following looks better, and text + capitalization are noticeably cleaner than Nano Banana 2 and Pro.
Some testers say it’s around 2x faster than GPT Image 2.5.
Overall, Nano Banana 2.1 feels like it’s all about speed — but I still think GPT Images delivers better quality.
👍246❤234🔥234🎉234