Holy Claude Sonnet 5.5 cooked this motion video
Replicated this motion design video in Claude Sonnet 5.5 in pure code with reference.
Results are good.
Used 10% of weekly limit.
Prompt below if you want to try and replicate ↓
some og you might have missed this in the gpt launch but MAI Code 2 looks so good in benchmarks, need to try it out.
Also its very cost efficient as well.
while working on ai motion design video i feel like they are either too fast or either too slow, this ruins the attention that video was meant to capture.
i feel sad looking at the current state of chinese ai labs, its been almost 2 months and their newest releases are hardly even coming close to opus 5.5. i might have to keep the existing claude subscription for much longer i gueess
@boringdev77 yeah i really hope so because the difference feels huge since i have subs for both the types. and the frontier performs way better with generous limits.
well if you were wondering why the Chinese labs have been silent lately, here you go: the 2.4 TRILLION param Qwen4-Max.
it actually beats Opus 5.5 on DeepSWE and Agents Last Exam in Qwen’s evals, and costs just $1.50/$4.50 per 1M input/output tokens, compared to Opus’s $4/$20.
it also falls behind substantially in some other benchmarks so take the numbers with a grain of salt,
but compared to Qwen3.8-Max, it’s a giant leap.
I just hope they drop a 27B soon too.
@shouchan099 the anti-distillation measures have been brutal, qwen took 2 months to rank even behind sonnet 5.5, they ain't even near opus 5.5 at any level tbh.
I just nominated @hi_mrinal for Engineering at the Cosign Awards. https://t.co/XTjncwWO1O
Can you all go and do it as well. following him in his footsteps. goated content 🙏
I think OpenAI is taking their time for cooking the new models, earlier they rushed to release em so the results were not so good but I am hoping the next models will amazing overall and we might see the toughest competition.