Thinking Machines ✓
@thinkymachines · joined 17 Feb 2025
Thinking, beeping, and booping. @tinkerapi
183 595Followers
1Following
71Posts total
19.6MViews on collected posts
Latest posts
Looking for new posts…
Cleaning data and aligning the reward function for RLVR takes expertise and effort upfront, but the result is a model that's state-of-the-art on a complex task.
Guest post by researchers at UIUC and Bridgewater, in collaboration with our team.
https://t.co/1UHTXFcZnW
75.5K views · 504 likes
· 41 reposts · 10 replies
Open on X →
LLMs with scaffolds have lagged on text-to-SQL, a task that relies on human judgment. By folding expert judgment into every part of RLVR on Tinker, @maxYuxuanZhu and @ddkang (UIUC and Bridgwater) trained the first text-to-SQL model to beat the human mark.
https://t.co/JxEQC31rX9
211.1K views · 363 likes
· 46 reposts · 5 replies
Open on X →
Try out Inkling and Inkling-Small on OpenRouter here:
https://t.co/R6UxnU8DJF
19.4K views · 135 likes
· 3 reposts · 0 replies
Open on X →
We want to improve Inkling’s agentic performance. To help us understand its real-world behavior, we are making it available for free on OpenRouter (only with agentic harnesses) for the next few weeks, starting now. We’ll use the data, disassociated from accounts, to better it.
348.3K views · 2.1K likes
· 126 reposts · 56 replies
Open on X →
@thinkymachines staged access is a lot of syllables for gatekeeping.
3.3K views · 41 likes
· 0 reposts · 1 replies
Open on X →
We’re also expanding our safety team. If you’re interested in joining, please reach out!
https://t.co/yS2F8gZqhh
20.5K views · 84 likes
· 1 reposts · 4 replies
Open on X →
@thinkymachines Congrats guys on the launch! Inkling is the first open 1T multimodal model that supports image, audio, and text. 🤯🔥
We uploaded Dynamic GGUFs for those who want to run Inkling locally: https://t.co/J1RJ5z5BBn
60.4K views · 580 likes
· 35 reposts · 14 replies
Open on X →
Inkling is the first in a family. We’ve included some details of Inkling-Small, a lighter-weight model trained on a similar recipe, with full weights to follow.
106.3K views · 497 likes
· 12 reposts · 6 replies
Open on X →
We hope you enjoy Inkling, and as always we’re keen to see what you build.
96.8K views · 470 likes
· 4 reposts · 10 replies
Open on X →
Cost and latency are important in real-world use cases. Inkling's continuous thinking effort lets you pick your point on the cost/performance curve—reaching the same score with a fraction of the tokens.
155.1K views · 703 likes
· 9 reposts · 3 replies
Open on X →
We’re grateful to our partners for their day-0 support across the open-source ecosystem: @togethercompute, @FireworksAI_HQ, @databricks, @UnslothAI, @modal, @baseten, @lightseekorg, @inferact on VLLM, @radixark on SGLang.
105.8K views · 589 likes
· 12 reposts · 7 replies
Open on X →
Inkling natively understands and reasons across text, audio, and images. It’s strong on audio in particular, ranking among the strongest open-weights models on VoiceBench, MMAU, and AudioMC.
120.4K views · 583 likes
· 12 reposts · 3 replies
Open on X →
Today, we are introducing Inkling.
Inkling reasons efficiently across text, image, and audio modalities. We are making the full weights available.
https://t.co/Ghebq5mG30
Available today for fine-tuning on Tinker. Play with it in the Inkling Playground. 🧵
8.3M views · 14.9K likes
· 2K reposts · 563 replies
Open on X →
@thinkymachines The blog is quite related to MiniLLM (https://t.co/4oVvlmDekr) in June 2023.
MiniLLM illustrated the mode-seeking behavior of reverse KL, designed a policy gradient method to optimize it, compared it to off-policy methods like SeqKD, and derived its relationship t
4.9K views · 36 likes
· 7 reposts · 2 replies
Open on X →
@thinkymachines Nice blog! It shares similar points with our previous DRP paper (https://t.co/bKOMgQ1tUC) that leverages a strong teacher to revise and polish the student model's reasoning trajectory rather than direct distillation for better learnability. https://t.co/RMeGQARnk8
10.1K views · 36 likes
· 5 reposts · 0 replies
Open on X →
@thinkymachines Is this the first time an AI lab have define the term mid-training means? https://t.co/DrS0L7La5P
9.6K views · 97 likes
· 1 reposts · 2 replies
Open on X →
Our latest post explores on-policy distillation, a training approach that unites the error-correcting relevance of RL with the reward density of SFT. When training it for math reasoning and as an internal chat assistant, we find that on-policy distillation can outperform other ht
1.9M views · 2.8K likes
· 402 reposts · 59 replies
Open on X →
@thinkymachines This most interesting link is broken 404: https://t.co/HbjtKHwZqA
84.6K views · 113 likes
· 1 reposts · 1 replies
Open on X →
Tinker advances our mission of enabling more people to do research on cutting-edge models and customize them to their needs.
https://t.co/raQkVTiUNB
70.3K views · 343 likes
· 18 reposts · 4 replies
Open on X →
Introducing Tinker: a flexible API for fine-tuning language models.
Write training loops in Python on your laptop; we'll run them on distributed GPUs.
Private beta starts today. We can't wait to see what researchers and developers build with cutting-edge open models! https://t.
4.2M views · 5.8K likes
· 782 reposts · 240 replies
Open on X →
@thinkymachines Breakdown of the blog:
Imagine you ask a friend the same question twice, and they give you slightly different answers each time. This is like what happens with big computer programs called Large Language Models, or LLMs. Even when you tell them to be super precis
12.9K views · 111 likes
· 17 reposts · 1 replies
Open on X →
@thinkymachines tldr: LLM inference nondeterminism isn't just floating-point non-associativity or GPU concurrent execution, the core culprit is batching variance, where server load unpredictably alters numerics. Batch-invariant kernels unlock true reproducibility, finally making
43.3K views · 386 likes
· 16 reposts · 7 replies
Open on X →
Today Thinking Machines Lab is launching our research blog, Connectionism. Our first blog post is “Defeating Nondeterminism in LLM Inference”
We believe that science is better when shared. Connectionism will cover topics as varied as our research is: from kernel numerics to http
3.5M views · 7.6K likes
· 1.2K reposts · 230 replies
Open on X →
Against accounts of the same size
5 posts from the last 90 days, next to the 100K–1M follower range. shown to more people than peers of the same size.
Median views75 493this account5 983median for 100K–1M
Reach, %41.12%this account1.49%median for 100K–1M
Engagement, %0.72%this account0.93%median for 100K–1M
| Metric | This account | Median for 100K–1M | Ratio |
|---|---|---|---|
| Median views per post | 75 493 | 5 983 | 12.6× |
| Reach (views ÷ followers) | 41.12% | 1.49% | 27.6× |
| Engagement rate | 0.72% | 0.93% | 0.77× |
Others in this range → Compare with another account → How these benchmarks are built →
Growth & engagement
How the posts we collected actually performed: views and reaction rate post by post, what the audience did with them, and where the follower count goes.
Views per post
8.3M15 Jul
120.4K
105.8K
155.1K
96.8K
106.3K
60.4K
20.5K31 Jul
3.3K1 Aug
348.3K21 Aug
19.4K
4.7K
211.1K27 Aug
75.5K
Last 14 collected posts, oldest on the left. The scale is logarithmic: one post can outrun the rest a hundred times over.
Engagement rate per post
0.22%15 Jul
0.50%
0.58%
0.46%
0.50%
0.49%
1.05%
0.43%31 Jul
1.27%1 Aug
0.67%21 Aug
0.72%
1.23%
0.20%27 Aug
0.74%
Reactions — likes, reposts, replies and quotes — divided by views. Median for 100K–1M accounts is 0.93%.
What the audience does
Likes58.6%39 252 in total
Reposts7.2%4 818 in total
Replies1.8%1 233 in total
Quotes2.7%1 800 in total
Bookmarks29.6%19 829 in total
Share of every reaction we collected for this account. Replies mean argument, reposts mean endorsement, bookmarks mean the post was worth keeping.
The follower curve appears once this account has two daily snapshots — we take one a day, and this one is on its first.
Similar accounts
✰‿❤⌒*✰ ᏘᏰᏰᎽ ✰‿❤⌒*✰
@1LA_Lady
367.2K followers
· 137.9K posts
Hiten Shah ✓
@hnshah
367.2K followers
· 127.2K posts
Carl Niehaus ✓
@niehaus_carl
367.2K followers
· 31.2K posts
Ideam Colombia ✓
@IDEAMColombia
367.1K followers
· 128.7K posts
Kanak Mani Dixit ✓
@KanakManiDixit
367.1K followers
· 53.7K posts
Christina Bobb ✓
@christina_bobb
367K followers
· 5K posts
UFC on TNT Sports ✓
@ufcontnt
367K followers
· 52.9K posts
Olivia Attwood Dack
@oliviajade_att
367K followers
· 4.8K posts