Sabitlenmiş Tweet
Tinker
121 posts
Tinker
@tinkerapi
I tink, therefore I am. Post-training API by @thinkymachines
San Francisco Katılım Ocak 2026
1 Takip Edilen6.1K Takipçiler
Tinker retweetledi
@ChenMoneyQ Thanks for the catch! The applications links should work now:
form.typeform.com/to/E9wVFZJJ
English
@tinkerapi Ah awesome! The application form seems to be closed tho... could you help generate a new link?
English
Kimi K2.6 from @Kimi_Moonshot and Qwen3.6-35B-A3B from @Alibaba_Qwen are now available on Tinker. Both models offer improvements in long-horizon agentic reliability over the previous versions, at two distinct points on the size-capability spectrum.
English
@Alibaba_Qwen We’re also adding Qwen3.6-27B, a dense model for thorough fine-tuning alongside the 35B-A3B MoE.
English
@ChenMoneyQ It is! We approve applications on a rolling basis, you can read more and apply here:
thinkingmachines.ai/news/tinker-re…
English
@tinkerapi Is it still possible to apply for Tinker research compute grant?
English
@HackingDave @HackingLZ @cantcomputer check out @tinkerapi — nice API for training your own models and then running them on your own hardware
English
Pulling the trigger on ordering 8xh100s for TrustedSec. The inconsistencies on frontier models plus how deep we are going with research is a must.
Now I’ll have my own dedicated coding system. Excited !
Maybe I’ll share with @HackingLZ and @cantcomputer ..
English
Exciting work from @wzenus, supported by Tinker grants!
Zihan "Zenus" Wang ✈️ ICLR@wzenus
In Agent RL, models suffer from Template Collapse. They generate vast, diverse outputs (High Entropy) that lose all meaningful connection to the input prompt (Low Mutual Information). In other words, agent learn different ways to say nothing. 🚀 Introducing RAGEN-v2 -- Here's how we define and fix such silent failure modes in Agent RL. 🧵
English
@skoshx hey Rasmus---sorry about this! We'd love to refund you for the credits you spent here and help out with getting your run over the line. Would you mind emailing tinker@thinkingmachines.ai with some more details of what you were trying to do?
English
@tinkerapi spent 100$ training on 247M tokens, and got ZERO checkpoints, because if you run out of usage on your account, progress isn't saved, the training run just fails and you lose ALL PROGRESS...
English
Tinker retweetledi
We built a new task to test AI research capabilities! Agents asked to use @tinkerapi from @thinkymachines to train a model on logic games. That involves writing full training pipeline, running experiments across recipes, and submitting the best model.
Proximal@ProximalHQ
FrontierSWE was built with collaborators from industry and academia to ensure that tasks are diverse and reflect real work engineers and researchers encounter. We specifically thank our partners @Modular, @PrimeIntellect and @thoughtfullab for their contributions
English
that's us!
Justus Mattern@MatternJustus
Another task tests AI research capabilities: using @tinkerapi from @thinkymachines, agents are asked to post-train an agent to play logic games, which involves writing an entire training pipeline and running experiments with different recipes to finally submit the best model
English
Coding agents are racing towards strong performance over long horizons. @ProximalHQ's FrontierSWE throws down a rigorous benchmark, and we're thrilled that Tinker gets to play a part!
Justus Mattern@MatternJustus
Introducing FrontierSWE, an ultra-long horizon coding benchmark. We test agents on some of the hardest technical tasks like optimizing a video rendering library or training a model to predict the quantum properties of molecules. Despite having 20 hours, they rarely succeed
English
@yacinelearning thanks Yacine! Pls let us know if you have any feedback or other things you'd like to see.
English
Tinker retweetledi
okay yes nice that’s the type of learning material I love to see
will def go through these
Tinker@tinkerapi
First, to get you started, we've created 23 tutorials to walk you from the API basics to advanced training techniques and deploying models into production. tinker-docs.thinkingmachines.ai/tutorials/
English
@wickedbrok Hi! We're finalising our Q2 roadmap and will let you know soon :)
English
@tinkerapi When will you guys add full fine tuning. It’s much needed for research comparisons.
English
please note we did not pay brydon to say this (we pay him to do research)
Brydon Eastman@brhydon
I know it's self serving to say, but man I would've killed for a resource like Tinker and the tutorials, the cookbook, etc back when I was in undergrad. Following @karpathy blogs and training RNNs on a crappy Acer *was* fun, but doing bigger things with less setup is such a boon
English
Tinker retweetledi
so many good tutorials on here would highly recommend checking it out if you haven't yet
Tinker@tinkerapi
We’ve redesigned our docs with easy access to SDK reference, tutorials, support, and our newly updated cookbook---v0.3.0! Whether you’re writing your first training loop in Tinker or debugging async RL, we want to make it easier to find what you need.
English
The full list of updates is in our changelog. We can’t wait to see what you build!
github.com/thinking-machi…
English