Seeing all the Opus 5.5 motion design videos lately, I had to try one too.
I gave Claude our site — https://t.co/0lk8S4KfyT — and asked it to turn the site into a 15-second launch film.
Here’s how it turned out 👇
All rendered from code: HTML → headless Chrome → FFmpeg.
Per task? Qwen was cheaper on 4 of 6 benchmarks. Just 7.7% lower overall (geometric mean).
Qwen / Kimi
Vibe Code $41.40 / $7.49 (5.5× the cost for +1.77 pts)
Tax Agent $1.03 / $3.64
Legal Research $2.59 / $7.23
Vals Index $9.33 / $6.44
https://t.co/ejs6MAWvhU
Qwen4-Max vs Kimi K3.1 Pro. Slight edge for Qwen. About half the blended token price.
Qwen / Kimi
Coding 64.6 / 64.0
Tool use 59.6 / 58.3
Qwen leads 3 of 3 professional work tests.
API per 1M tokens
Input $1.50 / $2
Output $4.50 / $10
Per task? Depends on the benchmark.
DeepSeek is opening up its built-in capabilities to users through an API. Plenty of devs want to build their own agents but are missing good tools like search, OCR, and other stuff. Now there’s a new choice—and DS’s tools are crazy cheap.
🚀 The full DeepSeek App experience, now via API
Search, files, links, safety, and voice. The capabilities behind the DeepSeek App are now available through the DeepSeek API.
Enable built-in features with a few parameters, or call standalone APIs such as OCR, ASR, and TTS. Build richer conversations or recreate the full DeepSeek App in your own application.
Try our open-source frontend demo, with support for importing DeepSeek conversation history.
🔗 Demo: https://t.co/TI0Iw5qLQK
📦 Code: https://t.co/ImA0LrgUgX
📖 Details: https://t.co/t0dJxXM6Lk
1/4
@thsottiaux If OpenAI is considering a $50 tier because ChatGPT Plus has much lower usage limits than Claude, then the $20 subscription may be pointless—at the same $20 price point, Claude already offers far more usage. Do you have better idea?
So if you’re a ChatGPT Plus subscriber, there seems to be no reason to keep using it now. Switching to the $20 Claude subscription is the best choice. Is there any reason to still use ChatGPT Plus?
Saw some confusions, but all Pro plans have no 5-hour limits, and give the following usage compared to Plus.
✅ Pro 100 = 10X
✅ Pro 200 = 20X
✅ Pro 500 = 50X, access to ultrafast
We compared Inkling 2 Small with DeepSeek V4.1 Flash across 23 benchmarks.
DeepSeek still wins 13–9 (1 tie), but Inkling takes 4/6 multimodal tests. That's pretty impressive.DeepSeek is still better at coding, though.
Full comparison:
https://t.co/TFryur6ijb
Looks like the US finally has an open-weight model that can actually compete with China's best.
DeepSeek and Qwen have been dominating for a while. OpenAI's GPT-OSS and Google's Gemma never really caught up.
But Thinking Machines' Inkling 2 Small might change that. 👀
The benchmarks Baidu presented for ERNIE 5.5 Flash are so outdated that they are essentially worthless. I wonder why they would use such benchmarks to compare the model with others. Don’t they realize that these benchmarks can no longer distinguish between models?
ERNIE 5.5 Flash Base is now open source — a 280B-parameter model with 14B active per token, and it already matches SOTA at its size!
And this is just the start — the full ERNIE 5.5 Flash is on its way. 👀
Here's another GIF generation test with GPT-Image-2.6. #GPTImage
Same prompt, different visual style. This time I went for a cleaner, more minimal look instead of the neon sci-fi aesthetic.
Which one do you prefer — GIF 1 or GIF 2?
1/2
OpenAI just dropped GPT-Image-2.6, and it can now generate animated GIFs!
I tried it in ChatGPT with a prompt to turn AI model chaos into a clean comparison dashboard.
Here's the 8-second GIF it generated.
What do you think? 👀
1/2
OpenAI just dropped GPT-Image-2.6, and it can now generate animated GIFs!
I tried it in ChatGPT with a prompt to turn AI model chaos into a clean comparison dashboard.
Here's the 8-second GIF it generated.
What do you think? 👀