tinygrad
Hacker ApprovedRanked 5 of 19 machine learning libraries by recommenders
Praised for a small, readable deep learning framework that competes with PyTorch in far less code and reaches many hardware backends. Common caveat: little production use yet and slow inference.
Praised for a small, readable deep learning framework that competes with PyTorch in far less code and reaches many hardware backends. Common caveat: little production use yet and slow inference.
56 people gave an opinion
- recommend it
- 40
- criticize it
- 16
- approval share
- 71%
Recommended in 9 of 13 months
- November 2025: recommended
- December 2025: recommended
- January 2026: recommended
- February 2026: recommended
- March 2026: recommended
- April 2026: recommended
- May 2026: recommended
- June 2026: not recommended
- July 2026: recommended
- August 2026: recommended
- September 2026: not recommended
- October 2026: not recommended
- November 2026: not recommended (month in progress)
Why it carries the badge
- 40 people recommend itNeeds 10 or more, met.
- 2.5 recommenders for every criticNeeds 2 to 1 or better, met.
- Recommended in 9 of 13 monthsNeeds 4 or more, met.
The same three rules apply to every product, and the badge can’t be bought. How the badge works
What people say about it
The reasons people give, with the number of comments that give each one. Open a reason to read what they said.
Why people recommend it
Each mark is one comment
Small, readable codebase
“There isn't any part of it that looks terribly hard to understand when compared to other code.”
Comment from 31 Dec 2025 in “Five Years of Tinygrad” “I don't think this is bad, if you know Python then most of the code will be fine for you.”
Comment from 31 Dec 2025 in “Five Years of Tinygrad” “A 1000x reduction in code size and matching/exceeding in perf is worth looking at.”
Comment from 30 Dec 2025 in “Five Years of Tinygrad” - Comment from 31 Dec 2025 in “Five Years of Tinygrad”
- Comment from 31 Dec 2025 in “Five Years of Tinygrad”
Runs on many hardware backends
“tinygrad does have pretty neat support for sharding things across various devices relatively easy”
Comment from 4 Apr 2026 in “Apple approves driver that lets Nvidia eGPUs work with Arm Macs” “the same stuff allows direct gpu communication so that even mac nvidia thing is possible”
Comment from 24 Dec 2025 in “Nvidia buying AI chip startup Groq for about $20B in cash” “He was able to run nvidia gpu on mac via thunderbolt with tinygrad.”
Comment from 29 Dec 2025 in “Five Years of Tinygrad” - Comment from 10 Apr 2026 in “Tinygrad reenables e-GPU support on Macs”
Direct path to hardware
“the most interesting thing about tinygrad was that theoretically you could render a model all the way into hardware”
Comment from 21 Mar 2026 in “Tinybox- offline AI device 120B parameters” “The TinyGrad approach of going straight to the hardware is telling”
Comment from 13 Apr 2026 in “Taking on CUDA with ROCm: 'One Step After Another'” “the push for a more vertical integration with tinygrad”
Comment from 22 Mar 2026 in “Tinybox- offline AI device 120B parameters” Competes with PyTorch in fewer lines
“The fact that it competes with PyTorch in so few lines speaks to the incredibly low incidental complexity imposed by Tinygrad.”
Comment from 29 Dec 2025 in “Five Years of Tinygrad” “If your needs can be expressed as tensor operations or neural network stuff that tinygrad supports, might as well use that”
Comment from 17 Feb 2026 in “BarraCUDA Open-source CUDA compiler targeting AMD GPUs”
What people warn about
Each mark is one comment
Dense, terse code hard to read
“clearly formatting decisions were made to keep line count down over the flow of code”
Comment from 31 Dec 2025 in “Five Years of Tinygrad” “one look at pm_reduce_collapse and I’m glad I don’t have to work in this codebase.”
Comment from 31 Dec 2025 in “Five Years of Tinygrad” “This operation still looks like a incomprehensible wall of code 40 years later”
Comment from 31 Dec 2025 in “Five Years of Tinygrad” - Comment from 30 Dec 2025 in “Five Years of Tinygrad”
- Comment from 30 Dec 2025 in “Five Years of Tinygrad”
- Comment from 31 Dec 2025 in “Five Years of Tinygrad”
Little proven production use
“I haven’t seen tinygrad used for any mainstream production project or thing of value, yet.”
Comment from 22 Mar 2026 in “Tinybox- offline AI device 120B parameters” “I haven't heard many (any?) war-stories about a TinyBox in production”
Comment from 20 Nov 2025 in “Why CUDA translation wont unlock AMD” Big claims, little to show
“Lots of words and weird analogies to say basically nothing.”
Comment from 29 Dec 2025 in “Five Years of Tinygrad” “They're claiming a lot, but don't have much to show for it”
Comment from 16 Aug 2026 in “Chestnut – eGPU dock with open-source firmware”
Other machine learning libraries people recommend
The four most recommended other machine learning libraries, with the same counts for each, so you can compare like for like. The tick on each bar marks the 2 to 1 line the badge needs.
See all 19 machine learning libraries, rankedWhat people said
40 of the 58 comments that mention it, newest first. Each links to the original comment.
“They're claiming a lot, but don't have much to show for it”
“things like TinyGrad offer a hint of what's possible”
“Absolutely amazing framework and you should absolutely check them out.”
“The TinyGrad approach of going straight to the hardware is telling”
“it has broader device support than ROCm”
“And this time it supports both Nvidia Ampere+ and Amds”
“tinygrad does have pretty neat support for sharding things across various devices relatively easy”
“the push for a more vertical integration with tinygrad”
“I haven’t seen tinygrad used for any mainstream production project or thing of value, yet.”
“the most interesting thing about tinygrad was that theoretically you could render a model all the way into hardware”
“If your needs can be expressed as tensor operations or neural network stuff that tinygrad supports, might as well use that”
“similar to what tinycorp is trying to do”
“tinygrad, and more specifically tinyJIT are super useful abstractions”
“Good code but on the way to terse”
“the code doing it is very clean and clear”
“clearly formatting decisions were made to keep line count down over the flow of code”
“This operation still looks like a incomprehensible wall of code 40 years later”
“I don't think this is bad, if you know Python then most of the code will be fine for you.”
“TBH that looks pretty decent to me. GPU programming is very very complicated.”
“I really hope tinygrad succeeds with their mission of commoditizing the petaflop.”
“There isn't any part of it that looks terribly hard to understand when compared to other code.”
“This is more readable than almost any large python codebase to me at first glance.”
“one look at pm_reduce_collapse and I’m glad I don’t have to work in this codebase.”
“Running a formatter on that codebase would probably at least triple it.”
“A 1000x reduction in code size and matching/exceeding in perf is worth looking at.”
“The mission is to commoditize the petaflop.”
“Not sure what's the point of this extreme code-golfing.”
“a true engineer running an ambitious very open company”
“it looks quite amusing I managed to run it right away and make fixes in one of the tasks”
“super cool project and I wish them the best”
“I think it has great potential for deployments on edge systems.”
“He was able to run nvidia gpu on mac via thunderbolt with tinygrad.”
“The fact that it competes with PyTorch in so few lines speaks to the incredibly low incidental complexity imposed by Tinygrad.”
“I find this organizational structure compelling”
“Lots of words and weird analogies to say basically nothing.”
“the same stuff allows direct gpu communication so that even mac nvidia thing is possible”
“Stuff like tinygrad will change this.”
“tinygrad’s small set of operations and laziness made it easy to implement”
“I haven't heard many (any?) war-stories about a TinyBox in production”
“is the only viable alternative to CUDA that I have seen popup in the past few years.”
Counts come from 1,795,640 Hacker News comments, November 2025 to November 2026. How we count. Report a mistake.