my biggest unconfirmed conspiracy theory is that when people complain about a model being "nerfed", it’s actually providers using heterogenous hardware (GPUs, TPUs, AMD, Inferentia) which produce varying output qualities. it lines up with nerfing happening shortly after launch or during on-peak when at capacity
Show more
what if we could simulate jev with an oss language model and token probabilities?
we break down how jev works and shipped a demo running with DiffusionGemma on @CloudflareDev workers ai:
Show more
which of your favorite token brokers are contributing back to open source ai? shall we count the ways?
a slice of data: number of PRs upstreamed and merged into the most popular oss inference engines, by PR contributor org
Show more
GLM 5.3 (the big one) from @Zai_org is also up on @CloudflareDev workers ai now! have fun on day 0 🧡
glm 5.3 flash from @Zai_org (fka ox alpha) is up on @CloudflareDev workers ai 🐮
‣ low cost, high intelligence
‣ 1M context window
‣ first multimodal model from the GLM family
‣ also accessible via CF AI Gateway and @OpenRouter
Show more
new alpha —
a top model is coming to workers ai
:o :x
deepseeks are now available on @CloudflareDev workers ai. first models with 1M context window!
it is very insane to be on the other side of this. like a whack-a-mole nightmare
when we host a model on workers ai, we have to consider if we're going to be able to serve a model economically + what demand looks like + what capacity we have.
testing models on @OpenRouter allows us to launch these things in market and gather data. if we have to tear a model down, your traffic gets routed to other providers. once we're confident in our deployment, then we bring it on platform for Workers AI users, where we commit to running the model for longer.
it's been a joy to work with openrouter team to enable us to experiment with models quickly 🧡
Show more
