It feels suspicious that MiMo-V2.6 Pro gets 46 in de index while DeepSeek-V4.1 (https://artificialanalysis.ai/models/deepseek-v4-1-flash) gets 39. According to the appendix at the bottom of https://mimo.xiaomi.com/mimo-v2-6 the deepseek model sometimes surpasses mimo and it's not so far behind in capabilities. A week ago opus 5 appeared 1 points ahead of fable 5 despite fable being a much smarter model (this has been corrected already)

All of the chinese labs have been overfitting on benchmark data to game the results for a while now - MiMo and Deepseek are not anywhere near frontier and mostly compete with models like Luna - which they are still worse than.

There isn't much compelling reason to use these unless you are just averse to giving money to openai/altman. A $20 codex sub gives you ~$150 of luna use per weekly limit, while there isn't any good subsidized options for chinese models at all (and the few who were subsidizing, like opencode, rugpulled by reducing monthly limit to $60 to $15 with no notice to users).

If you hit the $20 codex 5 hour limit wouldn't that be a good situation to switch to a Chinese model for a bit?

rugpull is a very strong word... I agree that they are not great with their pricing/credit communications, but they do state the multipliers quite clearly in various places. And plenty still have $60 or $30.

The main AA benchmark keeps changing, and had to be radically changed when Astra came out and showed zero improvement over GPT 5.6 Sol in their benchmark. Opus 5 is still 1 point ahead of Fable 5.0 on the index, if you manually add Fable 5.0 back into the list, so it hasn't actually been "corrected". It's only Fable 5.1 that is shown as ahead of Opus 5.

The AA benchmark is a weighted average of other benchmarks and some internal ones. I think the difficult part is finding benchmarks that reflect your own use of the models.

The way Artificial Analysis keeps changing their weights feels kind of like deciding who the winner should be and making the weights reflect that. They’ve been changing their weights to add more weight to improved long-running agentic capabilities, but doing so means they’re reducing the relative importance of world knowledge and of writing ability.

I’ll grant that maybe world knowledge isn’t that important for these models. But writing ability is important for human understanding, and I think the weird turns of phrase and word choices reflect the labs’ underweighting of the importance of human understanding.

I mean artificial is in their name....

> It feels suspicious that MiMo-V2.6 Pro gets 46 in de index while DeepSeek-V4.1 gets 39.

Why?

DS 4.1 is good but it’s clearly not as “smart” as non-flash models- it just doesn’t have the training data. Without a solid plan, it goes off the rails pretty regularly.

I installed this because I wanted the sqlite database that track when I'm turning each page

Very excited about this!

I wonder if these fusion techniques could help to run better local AI by streaming tokens from multiple machines and combining them

Sarah Cadigan-Fried has designed some very cool modern pixel art fonts worth checking! https://www.soft-type.com/

There's an interesting symmetry between the knitting, perler bead and pixel art crowds.

There was a talk at a Linux conference a while back relating knitting to programming and I’ve yet to watch it because the audio on YT wasn’t great but it’s on my list.

I find knitting very soothing, and it also scratches the same itch as programming.

Old looms could make patterns and it's considered by many that the way to "encode" those different patterns was actually an early of programming.

See also, beach pebbles.

There's also a growing interest on integrating DFlash: https://github.com/ggml-org/llama.cpp/issues/21978, I can't wait to see how it will compare against MTP

Yes, it's active and even had a new release last week

Makes me wonder if in the far future it will be ever possible to do something crazy like 7nm processes at home

The Titanic is pretty old. Could you build that in your backyard? The scale of the equipment required has drastically increased over time.

I wasn't able to pull some images and I lost 1h trying to diagnose network problems in my setup, but it didn't occur to me that "la liga" was the root cause . My workaround was to add "registry-mirrors": ["https://mirror.gcr.io"] in my /etc/docker/daemon.json

This is the correct answer.

Something that confused me for a while was the path "docker.io" used for pulling containers. There is not actually a container registry at "docker.io" - rather docker and podman are hard coded to convert it to either "registry-1.docker.io" or "index.docker.io".

Depending on your use case, you may not require a hosted registry at all. You could instead push images to your servers from wherever they're built.

I use a little script[0] to automate that when deploying some personal projects, but really it could be as simple as `docker save`/`scp`/`docker load` (especially for a one-off situation or when the images are small).

[0]: https://github.com/mkantor/docker-pushmi-pullyu

Thank you and your brother for all the amazing work, it's really inspiring to others <3

Thank you and appreciate it!