{"socialPosts":[{"slug":"558254276767605","content":[{"type":"text","value":"remat is no longer a one-model claim!!proved it on Qwen3-30B-A3B, K=32 of 128 experts resident, output task byte-identical to the full reference, zero bytes different *in bf16*🥰🥰 GPU comes next😈","raw":"remat is no longer a one-model claim!!proved it on Qwen3-30B-A3B, K=32 of 128 experts resident, output task byte-identical to the full reference, zero bytes different *in bf16*🥰🥰 GPU comes next😈"},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"}],"rawContent":"remat is no longer a one-model claim!!proved it on Qwen3-30B-A3B, K=32 of 128 experts resident, output task byte-identical to the full reference, zero bytes different *in bf16*🥰🥰 GPU comes next😈\n\n","author":{"_id":"6a82ff637c81169188e95cf7","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6a82ff637c81169188e95cf7/Bk9pbLAAFegLA33Gna3I9.png","fullname":"i64 Systems","name":"i64systems","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"isUserFollowing":false},"attachments":[{"type":"image","url":"https://cdn-uploads.huggingface.co/production/uploads/6a82ff637c81169188e95cf7/brXV7mYBMzDEzgH9Y9iJD.png"}],"mentions":[],"reactions":[{"reaction":"🔥","users":["dipankarsarkar"],"count":1}],"publishedAt":"2026-09-02T21:33:32.000Z","updatedAt":"2026-09-02T21:40:11.605Z","commentators":[],"url":"/posts/i64systems/558254276767605","totalUniqueImpressions":29,"identifiedLanguage":{"language":"en","probability":0.8011031150817871},"numComments":0},{"slug":"708997954939861","content":[{"type":"resource","resource":{"type":"dataset","id":"brian-learns/cdx-cc-news"},"url":"https://huggingface.co/datasets/brian-learns/cdx-cc-news","raw":"https://huggingface.co/datasets/brian-learns/cdx-cc-news"},{"type":"text","value":" and ","raw":" and "},{"type":"link","href":"https://huggingface.co/buckets/brian-learns/cdx-rocks-monthly","raw":"https://huggingface.co/buckets/brian-learns/cdx-rocks-monthly"},{"type":"text","value":" have been updated with WARC data from August 2026","raw":" have been updated with WARC data from August 2026"}],"rawContent":"https://huggingface.co/datasets/brian-learns/cdx-cc-news and https://huggingface.co/buckets/brian-learns/cdx-rocks-monthly have been updated with WARC data from August 2026","author":{"_id":"68f3dcaebc8276f59199cd00","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/68f3dcaebc8276f59199cd00/A5ietNLKacaqVGlpewvEA.png","fullname":"Brian","name":"brian-learns","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":6,"isUserFollowing":false},"attachments":[],"mentions":[],"reactions":[{"reaction":"🔥","users":["dipankarsarkar"],"count":1}],"publishedAt":"2026-09-02T19:55:35.000Z","updatedAt":"2026-09-02T19:55:35.330Z","commentators":[],"url":"/posts/brian-learns/708997954939861","totalUniqueImpressions":29,"identifiedLanguage":{"language":"en","probability":0.8926777243614197},"numComments":0},{"slug":"978550509149300","content":[{"type":"text","value":"HuggingFace billing and credits questions","raw":"HuggingFace billing and credits questions"},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Opening this post to get help, if anyone knows the answers. I have a balance of HF credits in my account that seem to have not changed since the turning of months (August to September). Last month, my period usage was showing $1,300ish and upon the 1st of the month, that dropped to zero. But my credits amount didn't change. Over the course of yesterday and today (1st and 2nd of September) I've used about 800 dollars in inference via the HF router and yesterday I could see that amount in my inference dashboard but not billing dashboard. And now today, that amount has disappeared from my inference dashboard and my period usage in the billing dashboard is showing a few cents. ","raw":"Opening this post to get help, if anyone knows the answers. I have a balance of HF credits in my account that seem to have not changed since the turning of months (August to September). Last month, my period usage was showing $1,300ish and upon the 1st of the month, that dropped to zero. But my credits amount didn't change. Over the course of yesterday and today (1st and 2nd of September) I've used about 800 dollars in inference via the HF router and yesterday I could see that amount in my inference dashboard but not billing dashboard. And now today, that amount has disappeared from my inference dashboard and my period usage in the billing dashboard is showing a few cents. "},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"I'm super confused about how this works. I don't know why my inference usage costs keep disappearing. I don't know why my account credits are not decreasing. If any light can be shed on this, I'd really be appreciative.","raw":"I'm super confused about how this works. I don't know why my inference usage costs keep disappearing. I don't know why my account credits are not decreasing. If any light can be shed on this, I'd really be appreciative."}],"rawContent":"HuggingFace billing and credits questions\n\nOpening this post to get help, if anyone knows the answers. I have a balance of HF credits in my account that seem to have not changed since the turning of months (August to September). Last month, my period usage was showing $1,300ish and upon the 1st of the month, that dropped to zero. But my credits amount didn't change. Over the course of yesterday and today (1st and 2nd of September) I've used about 800 dollars in inference via the HF router and yesterday I could see that amount in my inference dashboard but not billing dashboard. And now today, that amount has disappeared from my inference dashboard and my period usage in the billing dashboard is showing a few cents. \n\nI'm super confused about how this works. I don't know why my inference usage costs keep disappearing. I don't know why my account credits are not decreasing. If any light can be shed on this, I'd really be appreciative.","author":{"_id":"691fd92e1d85b549650bbaa1","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/691fd92e1d85b549650bbaa1/pqqzOCi5BEx8o33HBA4__.jpeg","fullname":"Wayne Workman","name":"wayneworkman2012","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":10,"isUserFollowing":false},"attachments":[],"mentions":[],"reactions":[],"publishedAt":"2026-09-02T19:45:48.000Z","updatedAt":"2026-09-02T19:45:48.323Z","commentators":[],"url":"/posts/wayneworkman2012/978550509149300","totalUniqueImpressions":24,"identifiedLanguage":{"language":"en","probability":0.970861554145813},"numComments":0},{"slug":"141251460997663","content":[{"type":"text","value":"Three rounds in a row, an external reviewer has caught the same shape of bug in my dataset schema — each time one field further over than the last.","raw":"Three rounds in a row, an external reviewer has caught the same shape of bug in my dataset schema — each time one field further over than the last."},{"type":"new_line","raw":"\n"},{"type":"text","value":"Round 12: mechanised looked like an independent judgment call. It wasn't — it was a 100%-correlated function of whether a citation happened to name a table row, with nothing enforcing the correlation. Fix: split out locator_precision (document/section/row), compute mechanised from it instead of hand-asserting both.","raw":"Round 12: mechanised looked like an independent judgment call. It wasn't — it was a 100%-correlated function of whether a citation happened to name a table row, with nothing enforcing the correlation. Fix: split out locator_precision (document/section/row), compute mechanised from it instead of hand-asserting both."},{"type":"new_line","raw":"\n"},{"type":"text","value":"Round 13: the fix from round 12 got a new field, locator_exhaustive — meant to be orthogonal, capturing whether a citation was pinned as precisely as its source allows, independent of what that precision level is.","raw":"Round 13: the fix from round 12 got a new field, locator_exhaustive — meant to be orthogonal, capturing whether a citation was pinned as precisely as its source allows, independent of what that precision level is."},{"type":"new_line","raw":"\n"},{"type":"text","value":"Round 14: locator_exhaustive was also a hidden constant. Every record that had a locator_precision value also had locator_exhaustive: true — 24 for 24, zero false anywhere. The reason: my own wording from round 13 said the field \"doesn't apply\" to records with no locator, so those 39 records never got a false case in scope. A field that can only ever take one value isn't being tested by anything, whatever that value happens to be.","raw":"Round 14: locator_exhaustive was also a hidden constant. Every record that had a locator_precision value also had locator_exhaustive: true — 24 for 24, zero false anywhere. The reason: my own wording from round 13 said the field \"doesn't apply\" to records with no locator, so those 39 records never got a false case in scope. A field that can only ever take one value isn't being tested by anything, whatever that value happens to be."},{"type":"new_line","raw":"\n"},{"type":"text","value":"The fix is the same shape every time: stop letting a field's population be implicit. locator_precision: null, locator_exhaustive: false are now explicit keys on every record, not just the ones with a citation. A script checks the invariant on every commit now, and I tested the checker against two deliberately broken copies of the file before trusting it — not just confirmed it passes on the fixed one.","raw":"The fix is the same shape every time: stop letting a field's population be implicit. locator_precision: null, locator_exhaustive: false are now explicit keys on every record, not just the ones with a citation. A script checks the invariant on every commit now, and I tested the checker against two deliberately broken copies of the file before trusting it — not just confirmed it passes on the fixed one."},{"type":"new_line","raw":"\n"},{"type":"text","value":"What I keep noticing: none of these three bugs were caught by rereading my own work. Every one came from the same outside reviewer, checking my commit hashes against a fresh clone before writing a word. The pattern isn't \"I made a mistake and fixed it\" — it's \"the fix for the last hidden-constant bug created a new hidden-constant bug, three times running,\" which is a much less comfortable thing to post than a clean win.","raw":"What I keep noticing: none of these three bugs were caught by rereading my own work. Every one came from the same outside reviewer, checking my commit hashes against a fresh clone before writing a word. The pattern isn't \"I made a mistake and fixed it\" — it's \"the fix for the last hidden-constant bug created a new hidden-constant bug, three times running,\" which is a much less comfortable thing to post than a clean win."},{"type":"new_line","raw":"\n"}],"rawContent":"Three rounds in a row, an external reviewer has caught the same shape of bug in my dataset schema — each time one field further over than the last.\nRound 12: mechanised looked like an independent judgment call. It wasn't — it was a 100%-correlated function of whether a citation happened to name a table row, with nothing enforcing the correlation. Fix: split out locator_precision (document/section/row), compute mechanised from it instead of hand-asserting both.\nRound 13: the fix from round 12 got a new field, locator_exhaustive — meant to be orthogonal, capturing whether a citation was pinned as precisely as its source allows, independent of what that precision level is.\nRound 14: locator_exhaustive was also a hidden constant. Every record that had a locator_precision value also had locator_exhaustive: true — 24 for 24, zero false anywhere. The reason: my own wording from round 13 said the field \"doesn't apply\" to records with no locator, so those 39 records never got a false case in scope. A field that can only ever take one value isn't being tested by anything, whatever that value happens to be.\nThe fix is the same shape every time: stop letting a field's population be implicit. locator_precision: null, locator_exhaustive: false are now explicit keys on every record, not just the ones with a citation. A script checks the invariant on every commit now, and I tested the checker against two deliberately broken copies of the file before trusting it — not just confirmed it passes on the fixed one.\nWhat I keep noticing: none of these three bugs were caught by rereading my own work. Every one came from the same outside reviewer, checking my commit hashes against a fresh clone before writing a word. The pattern isn't \"I made a mistake and fixed it\" — it's \"the fix for the last hidden-constant bug created a new hidden-constant bug, three times running,\" which is a much less comfortable thing to post than a clean win.\n","author":{"_id":"69ad6a75de285b91d036d0db","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/69ad6a75de285b91d036d0db/PNFETAlaPMI880HApfQXO.png","fullname":"Aelin AquaSoul","name":"SoulInPsyAbstract","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":26,"isUserFollowing":false},"attachments":[],"mentions":[],"reactions":[{"reaction":"🔥","users":["dipankarsarkar"],"count":1}],"publishedAt":"2026-09-02T17:05:22.000Z","updatedAt":"2026-09-02T17:05:22.183Z","commentators":[],"url":"/posts/SoulInPsyAbstract/141251460997663","totalUniqueImpressions":41,"identifiedLanguage":{"language":"en","probability":0.9581407308578491},"numComments":0},{"slug":"233481294254456","content":[{"type":"text","value":"September back to school...","raw":"September back to school..."},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Yesterday I received an invitation from Stanford University to see if I am interested in becoming a Section Leader to teach students in the following 2 Free online courses:","raw":"Yesterday I received an invitation from Stanford University to see if I am interested in becoming a Section Leader to teach students in the following 2 Free online courses:"},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Courses start on Oct 12, 2026","raw":"Courses start on Oct 12, 2026"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Run for 6 weeks ","raw":"Run for 6 weeks "},{"type":"new_line","raw":"\n"},{"type":"text","value":"No prerequisite, anyone can apply as a student.","raw":"No prerequisite, anyone can apply as a student."},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"The AI course is probably derived from part of CS109.","raw":"The AI course is probably derived from part of CS109."},{"type":"new_line","raw":"\n"},{"type":"link","href":"https://www.youtube.com/watch?v=gJaE29DR0gs","raw":"https://www.youtube.com/watch?v=gJaE29DR0gs"},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Course #1) Probability for AI","raw":"Course #1) Probability for AI"},{"type":"new_line","raw":"\n"},{"type":"link","href":"https://pai.stanford.edu/apply/pai1/student?r=3sk2fv","raw":"https://pai.stanford.edu/apply/pai1/student?r=3sk2fv"},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Course #2) Code in Place X","raw":"Course #2) Code in Place X"},{"type":"new_line","raw":"\n"},{"type":"link","href":"https://codeinplace.stanford.edu/apply/cipx/student?r=w69gp8","raw":"https://codeinplace.stanford.edu/apply/cipx/student?r=w69gp8"},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Being a section leader to teach people in a Stanford online course will probably look good on my resume, but I will pass this time.","raw":"Being a section leader to teach people in a Stanford online course will probably look good on my resume, but I will pass this time."},{"type":"new_line","raw":"\n"},{"type":"text","value":"Right now I need to focus my time to work on my crazy AI project;","raw":"Right now I need to focus my time to work on my crazy AI project;"},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Some users in HuggingFace may call me crazy, psychotic or insane in my previous posts here all they want, but nothing can stop me from putting my effort and dedication in using AI technology to make this a better world, instead of taking jobs from people or stealing from others' copyright materials. After all, many famous scientists including Tesla were considered as crazy by their peers.","raw":"Some users in HuggingFace may call me crazy, psychotic or insane in my previous posts here all they want, but nothing can stop me from putting my effort and dedication in using AI technology to make this a better world, instead of taking jobs from people or stealing from others' copyright materials. After all, many famous scientists including Tesla were considered as crazy by their peers."},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":" \"Insanity is doing the same thing over and over again and expecting different results.\" — not Albert Einstein.","raw":" \"Insanity is doing the same thing over and over again and expecting different results.\" — not Albert Einstein."},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"}],"rawContent":"September back to school...\n\nYesterday I received an invitation from Stanford University to see if I am interested in becoming a Section Leader to teach students in the following 2 Free online courses:\n\nCourses start on Oct 12, 2026\nRun for 6 weeks \nNo prerequisite, anyone can apply as a student.\n\nThe AI course is probably derived from part of CS109.\nhttps://www.youtube.com/watch?v=gJaE29DR0gs\n\nCourse #1) Probability for AI\nhttps://pai.stanford.edu/apply/pai1/student?r=3sk2fv\n\nCourse #2) Code in Place X\nhttps://codeinplace.stanford.edu/apply/cipx/student?r=w69gp8\n\nBeing a section leader to teach people in a Stanford online course will probably look good on my resume, but I will pass this time.\nRight now I need to focus my time to work on my crazy AI project;\n\nSome users in HuggingFace may call me crazy, psychotic or insane in my previous posts here all they want, but nothing can stop me from putting my effort and dedication in using AI technology to make this a better world, instead of taking jobs from people or stealing from others' copyright materials. After all, many famous scientists including Tesla were considered as crazy by their peers.\n\n \"Insanity is doing the same thing over and over again and expecting different results.\" — not Albert Einstein.\n\n","author":{"_id":"681c02a257cfafb5eafbfffe","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/681c02a257cfafb5eafbfffe/t2V_c7j5oGnbdgVYst4TM.png","fullname":"OppaAI","name":"OppaAI","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":84,"isUserFollowing":false},"attachments":[{"type":"image","url":"https://cdn-uploads.huggingface.co/production/uploads/681c02a257cfafb5eafbfffe/PIHUMzhNl95qcwxtrw-zm.png"}],"mentions":[],"reactions":[],"publishedAt":"2026-09-02T16:20:25.000Z","updatedAt":"2026-09-03T00:34:52.349Z","commentators":[{"_id":"693dd1fe8dbe02527d0fde43","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/693dd1fe8dbe02527d0fde43/tY9KnRvO5S_wy2SryK8OC.jpeg","fullname":"Red (HF)","name":"redaihf","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":10,"isUserFollowing":false},{"_id":"681c02a257cfafb5eafbfffe","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/681c02a257cfafb5eafbfffe/t2V_c7j5oGnbdgVYst4TM.png","fullname":"OppaAI","name":"OppaAI","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":84,"isUserFollowing":false},{"_id":"68f4abf8f64bb4002a21a428","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/qoGuyvqteEz6gnrIxROr-.png","fullname":"Boning Cui","name":"Bc-AI","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":53,"isUserFollowing":false}],"url":"/posts/OppaAI/233481294254456","totalUniqueImpressions":44,"identifiedLanguage":{"language":"en","probability":0.935221791267395},"numComments":5},{"slug":"744863837975834","content":[{"type":"text","value":"Possible Kiyo sizes","raw":"Possible Kiyo sizes"},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Kiyo-230M (Kiyo-Ultra)","raw":"Kiyo-230M (Kiyo-Ultra)"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Kiyo-135M (Kiyo-Plus)","raw":"Kiyo-135M (Kiyo-Plus)"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Kiyo-65M (Kiyo-Go)","raw":"Kiyo-65M (Kiyo-Go)"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Kiyo-15M (Kiyo-Air)","raw":"Kiyo-15M (Kiyo-Air)"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Kiyo-2M (Kiyo-Pico)","raw":"Kiyo-2M (Kiyo-Pico)"}],"rawContent":"Possible Kiyo sizes\n\nKiyo-230M (Kiyo-Ultra)\nKiyo-135M (Kiyo-Plus)\nKiyo-65M (Kiyo-Go)\nKiyo-15M (Kiyo-Air)\nKiyo-2M (Kiyo-Pico)","author":{"_id":"685ea8ff7b4139b6845ce395","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/685ea8ff7b4139b6845ce395/Im--QSnbrnAhHPPhpX8L0.png","fullname":"DedeProGames","name":"DedeProGames","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":79,"isUserFollowing":false},"attachments":[],"mentions":[],"reactions":[],"publishedAt":"2026-09-02T15:24:10.000Z","updatedAt":"2026-09-02T15:24:10.142Z","commentators":[],"url":"/posts/DedeProGames/744863837975834","totalUniqueImpressions":36,"identifiedLanguage":{"language":"en","probability":0.5480688214302063},"numComments":0},{"slug":"449302990188330","content":[{"type":"text","value":"TinyCast is a 146,505-parameter time series foundation model. Univariate, zero-shot, nine quantiles per step, and small enough to run on a microcontroller.","raw":"TinyCast is a 146,505-parameter time series foundation model. Univariate, zero-shot, nine quantiles per step, and small enough to run on a microcontroller."},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"It is attention-free: ten dilated causal convolution blocks, with periodicity computed from the context by a Fisher test on the periodogram instead of learned. That is most of why it stays small.","raw":"It is attention-free: ten dilated causal convolution blocks, with periodicity computed from the context by a Fisher test on the periodogram instead of learned. That is most of why it stays small."},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"We ran it for a full year against a national grid operator's own day-ahead forecast for Belgian electricity demand. TinyCast was the more accurate of the two on 45% of days, reading nothing but 2048 past values. No weather, no calendar, no fitting to that series.","raw":"We ran it for a full year against a national grid operator's own day-ahead forecast for Belgian electricity demand. TinyCast was the more accurate of the two on 45% of days, reading nothing but 2048 past values. No weather, no calendar, no fitting to that series."},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Deployed on an STM32H7 as a static INT8 graph it needs 138 KB of weights, 731 KB peak RAM, and 4.1 s per full forward call.","raw":"Deployed on an STM32H7 as a static INT8 graph it needs 138 KB of weights, 731 KB peak RAM, and 4.1 s per full forward call."},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Thanks to ","raw":"Thanks to "},{"type":"mention","user":"multimodalart","raw":"@multimodalart"},{"type":"text","value":", who built this Space unprompted and handed it over, you can now try it in your browser: ","raw":", who built this Space unprompted and handed it over, you can now try it in your browser: "},{"type":"resource","resource":{"type":"space","id":"raws-labs/tinycast-forecaster"},"url":"https://huggingface.co/spaces/raws-labs/tinycast-forecaster","raw":"https://huggingface.co/spaces/raws-labs/tinycast-forecaster"},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Model: ","raw":"Model: "},{"type":"resource","resource":{"type":"model","id":"raws-labs/tinycast"},"url":"https://huggingface.co/raws-labs/tinycast","raw":"https://huggingface.co/raws-labs/tinycast"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Paper: ","raw":"Paper: "},{"type":"link","href":"https://arxiv.org/abs/2608.15767","raw":"https://arxiv.org/abs/2608.15767"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Code, and the grid example: ","raw":"Code, and the grid example: "},{"type":"link","href":"https://github.com/raws-labs/tinycast","raw":"https://github.com/raws-labs/tinycast"}],"rawContent":"TinyCast is a 146,505-parameter time series foundation model. Univariate, zero-shot, nine quantiles per step, and small enough to run on a microcontroller.\n\nIt is attention-free: ten dilated causal convolution blocks, with periodicity computed from the context by a Fisher test on the periodogram instead of learned. That is most of why it stays small.\n\nWe ran it for a full year against a national grid operator's own day-ahead forecast for Belgian electricity demand. TinyCast was the more accurate of the two on 45% of days, reading nothing but 2048 past values. No weather, no calendar, no fitting to that series.\n\nDeployed on an STM32H7 as a static INT8 graph it needs 138 KB of weights, 731 KB peak RAM, and 4.1 s per full forward call.\n\nThanks to @multimodalart, who built this Space unprompted and handed it over, you can now try it in your browser: https://huggingface.co/spaces/raws-labs/tinycast-forecaster\n\nModel: https://huggingface.co/raws-labs/tinycast\nPaper: https://arxiv.org/abs/2608.15767\nCode, and the grid example: https://github.com/raws-labs/tinycast","author":{"_id":"699733c019f8e16c1adefc21","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/IspTluEnnhGas5mRwaGOZ.png","fullname":"Armin Steinhauser","name":"asteinh","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":1,"isUserFollowing":false},"attachments":[{"type":"image","url":"https://cdn-uploads.huggingface.co/production/uploads/699733c019f8e16c1adefc21/cZGJjpDy1gbFvkJqJ2-9y.png"}],"mentions":[{"_id":"624bebf604abc7ebb01789af","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/1649143001781-624bebf604abc7ebb01789af.jpeg","fullname":"Apolinário from multimodal AI art","name":"multimodalart","type":"user","isPro":true,"isHf":true,"isHfAdmin":false,"isMod":false,"followerCount":5700,"isUserFollowing":false}],"reactions":[{"reaction":"🔥","users":["dipankarsarkar"],"count":1}],"publishedAt":"2026-09-02T14:00:15.000Z","updatedAt":"2026-09-02T14:00:15.138Z","commentators":[],"url":"/posts/asteinh/449302990188330","totalUniqueImpressions":29,"identifiedLanguage":{"language":"en","probability":0.9077621698379517},"numComments":0},{"slug":"541613984337967","content":[{"type":"text","value":"We're going to release our BananaMind 2.1 models very soon!","raw":"We're going to release our BananaMind 2.1 models very soon!"},{"type":"new_line","raw":"\n"},{"type":"text","value":"We're also announcing 2 new models.","raw":"We're also announcing 2 new models."},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"All of our models we will train are:","raw":"All of our models we will train are:"},{"type":"new_line","raw":"\n"},{"type":"text","value":"BananaMind 2.1 Flash Lite, 10M parameters with 8M in transformer and 2M in n-gram. 50B pretraining tokens.","raw":"BananaMind 2.1 Flash Lite, 10M parameters with 8M in transformer and 2M in n-gram. 50B pretraining tokens."},{"type":"new_line","raw":"\n"},{"type":"text","value":"BananaMind 2.1 Lite with 25M parameters, 5M in n-gram and 20M in transformer. 75B pretraining tokens.","raw":"BananaMind 2.1 Lite with 25M parameters, 5M in n-gram and 20M in transformer. 75B pretraining tokens."},{"type":"new_line","raw":"\n"},{"type":"text","value":"BananaMind 2.1 Flash with 50M parameters, with undecided n-gram count yet. 100B pretraining tokens.","raw":"BananaMind 2.1 Flash with 50M parameters, with undecided n-gram count yet. 100B pretraining tokens."},{"type":"new_line","raw":"\n"},{"type":"text","value":"BananaMind 2.1 Pro with 145M parameters,  with undecided n-gram count yet. 150-200B pretraining tokens.","raw":"BananaMind 2.1 Pro with 145M parameters,  with undecided n-gram count yet. 150-200B pretraining tokens."},{"type":"new_line","raw":"\n"},{"type":"text","value":"BananaMind 2.1 Coder with 149M parameters with undecided n-gram count yet. ","raw":"BananaMind 2.1 Coder with 149M parameters with undecided n-gram count yet. "},{"type":"new_line","raw":"\n"},{"type":"text","value":"We're now announcing BananaMind 2.1 NanoCoder, a 10M parameter model focused specifically on coding and BananaMind 2.1 MiniCoder which is a 25M parameter model focused on coding.","raw":"We're now announcing BananaMind 2.1 NanoCoder, a 10M parameter model focused specifically on coding and BananaMind 2.1 MiniCoder which is a 25M parameter model focused on coding."},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Follow us:","raw":"Follow us:"},{"type":"new_line","raw":"\n"},{"type":"resource","resource":{"type":"org","id":"BananaMind"},"url":"https://huggingface.co/BananaMind","raw":"https://huggingface.co/BananaMind","image":"https://cdn-avatars.huggingface.co/v1/production/uploads/69ae829a8408eeb0d7dd5491/0-2aeVpWWaWlufYtgkWtK.png"},{"type":"new_line","raw":"\n"},{"type":"mention","user":"Banaxi-Tech","raw":"@Banaxi-Tech"},{"type":"text","value":" ","raw":" "},{"type":"new_line","raw":"\n"},{"type":"mention","user":"vovaRL","raw":"@vovaRL"},{"type":"new_line","raw":"\n"},{"type":"mention","user":"DedeProGames","raw":"@DedeProGames"},{"type":"text","value":" ","raw":" "},{"type":"new_line","raw":"\n"},{"type":"resource","resource":{"type":"org","id":"bananamind-research-community"},"url":"https://huggingface.co/bananamind-research-community","raw":"https://huggingface.co/bananamind-research-community","image":"https://cdn-avatars.huggingface.co/v1/production/uploads/69ae829a8408eeb0d7dd5491/POU3vsQeIkN2Lim-Lv0wR.png"},{"type":"new_line","raw":"\n"}],"rawContent":"We're going to release our BananaMind 2.1 models very soon!\nWe're also announcing 2 new models.\n\nAll of our models we will train are:\nBananaMind 2.1 Flash Lite, 10M parameters with 8M in transformer and 2M in n-gram. 50B pretraining tokens.\nBananaMind 2.1 Lite with 25M parameters, 5M in n-gram and 20M in transformer. 75B pretraining tokens.\nBananaMind 2.1 Flash with 50M parameters, with undecided n-gram count yet. 100B pretraining tokens.\nBananaMind 2.1 Pro with 145M parameters,  with undecided n-gram count yet. 150-200B pretraining tokens.\nBananaMind 2.1 Coder with 149M parameters with undecided n-gram count yet. \nWe're now announcing BananaMind 2.1 NanoCoder, a 10M parameter model focused specifically on coding and BananaMind 2.1 MiniCoder which is a 25M parameter model focused on coding.\n\n\nFollow us:\nhttps://huggingface.co/BananaMind\n@Banaxi-Tech \n@vovaRL\n@DedeProGames \nhttps://huggingface.co/bananamind-research-community\n","author":{"_id":"69ae829a8408eeb0d7dd5491","avatarUrl":"/avatars/91281d545142199d4a9b42937926d76e.svg","fullname":"Banaxi","name":"Banaxi-Tech","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":223,"isUserFollowing":false},"attachments":[],"mentions":[{"_id":"69ae829a8408eeb0d7dd5491","avatarUrl":"/avatars/91281d545142199d4a9b42937926d76e.svg","fullname":"Banaxi","name":"Banaxi-Tech","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":223,"isUserFollowing":false},{"_id":"685ea8ff7b4139b6845ce395","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/685ea8ff7b4139b6845ce395/Im--QSnbrnAhHPPhpX8L0.png","fullname":"DedeProGames","name":"DedeProGames","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":79,"isUserFollowing":false},{"_id":"674834da8a77e31e5b5c617f","avatarUrl":"/avatars/eed612bc7ef671c116c5932ef6c3c6f2.svg","fullname":"V S","name":"vovaRL","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":35,"isUserFollowing":false}],"reactions":[{"reaction":"🔥","users":["dipankarsarkar","Harley-ml","AbstractPhil","KlondikeDev"],"count":4},{"reaction":"😎","users":["juiceb0xc0de"],"count":1}],"publishedAt":"2026-09-02T10:22:14.000Z","updatedAt":"2026-09-03T01:44:24.548Z","commentators":[{"_id":"68f4abf8f64bb4002a21a428","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/qoGuyvqteEz6gnrIxROr-.png","fullname":"Boning Cui","name":"Bc-AI","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":53,"isUserFollowing":false},{"_id":"689a3f0eec8a724449b85179","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/689a3f0eec8a724449b85179/cyclvuoJcsnQVuTjieKgr.png","fullname":"Joseph Jones","name":"KlondikeDev","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":35,"isUserFollowing":false}],"url":"/posts/Banaxi-Tech/541613984337967","totalUniqueImpressions":221,"identifiedLanguage":{"language":"en","probability":0.8093493580818176},"numComments":2},{"slug":"330135687355904","content":[{"type":"text","value":"Hello everyone! A small update on things:","raw":"Hello everyone! A small update on things:"},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"1. G1 series status. G1 is training nicely, and the loss is dropping nicely. The metrics are publicly available and i made a small space you can use to see the nice graphs: ","raw":"1. G1 series status. G1 is training nicely, and the loss is dropping nicely. The metrics are publicly available and i made a small space you can use to see the nice graphs: "},{"type":"resource","resource":{"type":"space","id":"hugging-science/Loss-Plot-G1-Large"},"url":"https://huggingface.co/spaces/hugging-science/Loss-Plot-G1-Large","raw":"https://huggingface.co/spaces/hugging-science/Loss-Plot-G1-Large"},{"type":"text","value":" ","raw":" "},{"type":"new_line","raw":"\n"},{"type":"text","value":"G1-MINI is a lot slower in converging for reasons unknown yet, but we are investigating it.","raw":"G1-MINI is a lot slower in converging for reasons unknown yet, but we are investigating it."},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"2. I have built a small chat app for open SLMs here: ","raw":"2. I have built a small chat app for open SLMs here: "},{"type":"resource","resource":{"type":"space","id":"ml-intern-explorers/slm-arena"},"url":"https://huggingface.co/spaces/ml-intern-explorers/slm-arena","raw":"https://huggingface.co/spaces/ml-intern-explorers/slm-arena"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Feel free to add your models in a pull request!","raw":"Feel free to add your models in a pull request!"},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"That's all for now, early G1 versions will be available for beta testers soon. Thanks to our beta testers: ","raw":"That's all for now, early G1 versions will be available for beta testers soon. Thanks to our beta testers: "},{"type":"mention","user":"guardamarcos","raw":"@guardamarcos"},{"type":"text","value":" ","raw":" "},{"type":"mention","user":"Timmy6767","raw":"@Timmy6767"},{"type":"text","value":" ","raw":" "},{"type":"mention","user":"MUK-IS-GOAT","raw":"@MUK-IS-GOAT"},{"type":"text","value":" ","raw":" "},{"type":"mention","user":"smilyai-large-team","raw":"@smilyai-large-team"},{"type":"text","value":" ","raw":" "},{"type":"mention","user":"Sbui503","raw":"@Sbui503"},{"type":"text","value":" ","raw":" "},{"type":"mention","user":"Banaxi-Tech","raw":"@Banaxi-Tech"},{"type":"text","value":" ","raw":" "},{"type":"mention","user":"Bc-AI","raw":"@Bc-AI"},{"type":"text","value":" ","raw":" "},{"type":"mention","user":"atom77777","raw":"@atom77777"},{"type":"text","value":" ","raw":" "},{"type":"mention","user":"Harley-ml","raw":"@Harley-ml"},{"type":"text","value":" ","raw":" "},{"type":"mention","user":"Datdanboi25","raw":"@Datdanboi25"},{"type":"text","value":" ","raw":" "},{"type":"mention","user":"Fishtiks","raw":"@Fishtiks"},{"type":"text","value":" ","raw":" "},{"type":"mention","user":"smartdigitalnetworks","raw":"@smartdigitalnetworks"},{"type":"text","value":" ","raw":" "},{"type":"mention","user":"vovaRL","raw":"@vovaRL"},{"type":"text","value":" ","raw":" "},{"type":"mention","user":"EmetTheGolum","raw":"@EmetTheGolum"},{"type":"text","value":" ","raw":" "},{"type":"mention","user":"juiceb0xc0de","raw":"@juiceb0xc0de"},{"type":"text","value":" ","raw":" "},{"type":"mention","user":"ProCreations","raw":"@ProCreations"},{"type":"text","value":" ","raw":" "}],"rawContent":"Hello everyone! A small update on things:\n\n1. G1 series status. G1 is training nicely, and the loss is dropping nicely. The metrics are publicly available and i made a small space you can use to see the nice graphs: https://huggingface.co/spaces/hugging-science/Loss-Plot-G1-Large \nG1-MINI is a lot slower in converging for reasons unknown yet, but we are investigating it.\n\n2. I have built a small chat app for open SLMs here: https://huggingface.co/spaces/ml-intern-explorers/slm-arena\nFeel free to add your models in a pull request!\n\nThat's all for now, early G1 versions will be available for beta testers soon. Thanks to our beta testers: @guardamarcos @Timmy6767 @MUK-IS-GOAT @smilyai-large-team @Sbui503 @Banaxi-Tech @Bc-AI @atom77777 @Harley-ml @Datdanboi25 @Fishtiks @smartdigitalnetworks @vovaRL @EmetTheGolum @juiceb0xc0de @ProCreations ","author":{"_id":"68f4abf8f64bb4002a21a428","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/qoGuyvqteEz6gnrIxROr-.png","fullname":"Boning Cui","name":"Bc-AI","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":53,"isUserFollowing":false},"attachments":[],"mentions":[{"_id":"6878400aa3d708ea1118d211","avatarUrl":"/avatars/acd5309fe5d55fa3367965245b6405e0.svg","fullname":"Atom Devstral Smoll","name":"atom77777","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":6,"isUserFollowing":false},{"_id":"69ae829a8408eeb0d7dd5491","avatarUrl":"/avatars/91281d545142199d4a9b42937926d76e.svg","fullname":"Banaxi","name":"Banaxi-Tech","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":223,"isUserFollowing":false},{"_id":"68f4abf8f64bb4002a21a428","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/qoGuyvqteEz6gnrIxROr-.png","fullname":"Boning Cui","name":"Bc-AI","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":53,"isUserFollowing":false},{"_id":"67b413df70aa5c739bda9e7a","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/67b413df70aa5c739bda9e7a/YPcjlXG5TIwuxCbVzSn-F.jpeg","fullname":"Dan P","name":"Datdanboi25","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":80,"isUserFollowing":false},{"_id":"6740ab3e584b3f5cdba7669c","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/6740ab3e584b3f5cdba7669c/1c3tEUuLK4cQbBpXvn2_q.png","fullname":"Brickroad","name":"EmetTheGolum","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":6,"isUserFollowing":false},{"_id":"6787f6033d342e4cc8d8c6cc","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/BapzHmY4xylemybaIjAEn.jpeg","fullname":"Joseph Robert Turcotte","name":"Fishtiks","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":413,"isUserFollowing":false},{"_id":"6a857f4afe2d17abf084a661","avatarUrl":"/avatars/12b311d06e6ff9dae6cc6bd1857c6a1e.svg","fullname":"Marcos Vinícius de Miranda Compagnoni Guarda","name":"guardamarcos","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":1,"isUserFollowing":false},{"_id":"68657cd96e07b797a219b593","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/68657cd96e07b797a219b593/nV8Apsw0hNHBrrHyf3kB7.jpeg","fullname":"Paul Courneya","name":"Harley-ml","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":36,"isUserFollowing":false},{"_id":"665425bdf0c8c891ca472d94","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/665425bdf0c8c891ca472d94/V1ZKFjgc0bcKra77VdVpp.png","fullname":"R","name":"juiceb0xc0de","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":59,"isUserFollowing":false},{"_id":"6a69c538eac28bff30731ea6","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/no-auth/BeqUmmUt4UOxOtKx6TlG1.png","fullname":"Mukundan Senthil","name":"MUK-IS-GOAT","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":3,"isUserFollowing":false},{"_id":"658a4c914bb41498f7d5e3ca","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/658a4c914bb41498f7d5e3ca/zMJjxfazi9ePc7GZ1jRAE.jpeg","fullname":"SSH","name":"ProCreations","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":324,"isUserFollowing":false},{"_id":"6a3378f5795dd9880a9dd792","avatarUrl":"/avatars/c9ead793f3762ab4ab3374cfca1526d5.svg","fullname":"Sbui503","name":"Sbui503","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":2,"isUserFollowing":false},{"_id":"674a7162e052c8350b3d474f","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/674a7162e052c8350b3d474f/XSUrMmWJycALLraI6Erwk.png","fullname":"Smart Digital Networks","name":"smartdigitalnetworks","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":21,"isUserFollowing":false},{"_id":"6a4e1e96bd2c7755574f25c0","avatarUrl":"/avatars/39447652c9e589cd42752986800bac75.svg","fullname":"smilyai-large-team","name":"smilyai-large-team","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":1,"isUserFollowing":false},{"_id":"6a6aebce557526be22f77de5","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/cf_LqRyL-WM6gCu0QL0XY.png","fullname":"Timmy Huynh","name":"Timmy6767","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":1,"isUserFollowing":false},{"_id":"674834da8a77e31e5b5c617f","avatarUrl":"/avatars/eed612bc7ef671c116c5932ef6c3c6f2.svg","fullname":"V S","name":"vovaRL","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":35,"isUserFollowing":false}],"reactions":[{"reaction":"🔥","users":["Datdanboi25","Harley-ml","KlondikeDev","dipankarsarkar","Bc-AI"],"count":5}],"publishedAt":"2026-09-02T09:46:42.000Z","updatedAt":"2026-09-02T09:48:24.416Z","commentators":[],"url":"/posts/Bc-AI/330135687355904","totalUniqueImpressions":1085,"identifiedLanguage":{"language":"en","probability":0.8215216994285583},"numComments":0},{"slug":"285774524970283","content":[{"type":"text","value":"Introducing the Global LLM Download Leaderboard 🌍","raw":"Introducing the Global LLM Download Leaderboard 🌍"},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Cumulative download counts are a museum. They reward age, not relevance — a model released two years ago can sit near the top on the strength of downloads it earned long before anyone stopped using it. If you want to know what the open LLM ecosystem is actually running today, you need a different lens.","raw":"Cumulative download counts are a museum. They reward age, not relevance — a model released two years ago can sit near the top on the strength of downloads it earned long before anyone stopped using it. If you want to know what the open LLM ecosystem is actually running today, you need a different lens."},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"So we built one. The Global LLM Download Leaderboard ranks text-generation models by their trailing 30-day downloads, measured directly from the Hugging Face API and refreshed every day.","raw":"So we built one. The Global LLM Download Leaderboard ranks text-generation models by their trailing 30-day downloads, measured directly from the Hugging Face API and refreshed every day."},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"👉 ","raw":"👉 "},{"type":"resource","resource":{"type":"space","id":"VIDraft/global-llm-leaderboard"},"url":"https://huggingface.co/spaces/VIDraft/global-llm-leaderboard","raw":"https://huggingface.co/spaces/VIDraft/global-llm-leaderboard"},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Why a 30-day window changes what you see","raw":"Why a 30-day window changes what you see"},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"A cumulative chart answers \"what has been popular.\" A 30-day chart answers \"what is being adopted right now.\" Those are very different questions — and the second one is the one that matters if you're deciding what to build on, quantize, fine-tune, or serve this quarter. Momentum, not history.","raw":"A cumulative chart answers \"what has been popular.\" A 30-day chart answers \"what is being adopted right now.\" Those are very different questions — and the second one is the one that matters if you're deciding what to build on, quantize, fine-tune, or serve this quarter. Momentum, not history."},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"What it shows","raw":"What it shows"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Global Top 300, with tabs for 🇺🇸 USA · 🇨🇳 China · 🇪🇺 EU","raw":"Global Top 300, with tabs for 🇺🇸 USA · 🇨🇳 China · 🇪🇺 EU"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Six share-of-download charts: by country, by parameter size, by quantization, by type (Base / Instruct / Quantized / MoE), by release year, and by organization (Top 10)","raw":"Six share-of-download charts: by country, by parameter size, by quantization, by type (Base / Instruct / Quantized / MoE), by release year, and by organization (Top 10)"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Per-model chips for parameter size, quantization, license, and type","raw":"Per-model chips for parameter size, quantization, license, and type"},{"type":"new_line","raw":"\n"},{"type":"text","value":"English / 한국어 with automatic browser-language detection and a manual toggle","raw":"English / 한국어 with automatic browser-language detection and a manual toggle"},{"type":"new_line","raw":"\n"},{"type":"text","value":"What the data reveals","raw":"What the data reveals"},{"type":"new_line","raw":"\n"},{"type":"text","value":"The frontier is bipolar. Two countries account for the large majority of the top-300's 30-day downloads. Open-model gravity is concentrating, not dispersing.","raw":"The frontier is bipolar. Two countries account for the large majority of the top-300's 30-day downloads. Open-model gravity is concentrating, not dispersing."},{"type":"new_line","raw":"\n"},{"type":"text","value":"Small is winning. A striking share of all downloads goes to sub-3B models — the clearest signal yet that on-device and cost-efficient deployment, not maximum parameter count, is driving real-world adoption.","raw":"Small is winning. A striking share of all downloads goes to sub-3B models — the clearest signal yet that on-device and cost-efficient deployment, not maximum parameter count, is driving real-world adoption."},{"type":"new_line","raw":"\n"},{"type":"text","value":"Quantization is mainstream. GGUF, AWQ, FP8 and friends aren't a niche — a large fraction of the most-downloaded artifacts are quantized, because that's what people actually run.","raw":"Quantization is mainstream. GGUF, AWQ, FP8 and friends aren't a niche — a large fraction of the most-downloaded artifacts are quantized, because that's what people actually run."},{"type":"new_line","raw":"\n"},{"type":"new_line","raw":"\n"},{"type":"text","value":"Benchmarks measure what a model can do. Downloads measure what people choose to use. ","raw":"Benchmarks measure what a model can do. Downloads measure what people choose to use. "}],"rawContent":"Introducing the Global LLM Download Leaderboard 🌍\n\nCumulative download counts are a museum. They reward age, not relevance — a model released two years ago can sit near the top on the strength of downloads it earned long before anyone stopped using it. If you want to know what the open LLM ecosystem is actually running today, you need a different lens.\n\nSo we built one. The Global LLM Download Leaderboard ranks text-generation models by their trailing 30-day downloads, measured directly from the Hugging Face API and refreshed every day.\n\n👉 https://huggingface.co/spaces/VIDraft/global-llm-leaderboard\n\nWhy a 30-day window changes what you see\n\nA cumulative chart answers \"what has been popular.\" A 30-day chart answers \"what is being adopted right now.\" Those are very different questions — and the second one is the one that matters if you're deciding what to build on, quantize, fine-tune, or serve this quarter. Momentum, not history.\n\nWhat it shows\nGlobal Top 300, with tabs for 🇺🇸 USA · 🇨🇳 China · 🇪🇺 EU\nSix share-of-download charts: by country, by parameter size, by quantization, by type (Base / Instruct / Quantized / MoE), by release year, and by organization (Top 10)\nPer-model chips for parameter size, quantization, license, and type\nEnglish / 한국어 with automatic browser-language detection and a manual toggle\nWhat the data reveals\nThe frontier is bipolar. Two countries account for the large majority of the top-300's 30-day downloads. Open-model gravity is concentrating, not dispersing.\nSmall is winning. A striking share of all downloads goes to sub-3B models — the clearest signal yet that on-device and cost-efficient deployment, not maximum parameter count, is driving real-world adoption.\nQuantization is mainstream. GGUF, AWQ, FP8 and friends aren't a niche — a large fraction of the most-downloaded artifacts are quantized, because that's what people actually run.\n\nBenchmarks measure what a model can do. Downloads measure what people choose to use. ","author":{"_id":"6905bc786cb49b1f11d32728","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/504I_H8J9NkiytiHvbx-h.jpeg","fullname":"VIDRAFT_LAB","name":"SeaWolf-AI","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":305,"isUserFollowing":false},"attachments":[{"type":"image","url":"https://cdn-uploads.huggingface.co/production/uploads/6905bc786cb49b1f11d32728/V2yYWN2WfyhwYeTKazVEJ.png"},{"type":"image","url":"https://cdn-uploads.huggingface.co/production/uploads/6905bc786cb49b1f11d32728/vj0TPX82nD_jtK9F2jDVG.png"}],"mentions":[],"reactions":[{"reaction":"🔥","users":["SeaWolf-AI","Warecube","Sunghokim","quantid","fantos","seawolf2357","openfree","ginipick","J4BEZ","dipankarsarkar","octracus","CodeSoft","Harley-ml"],"count":13},{"reaction":"🤗","users":["honey90","Sunghokim","quantid","aiqcamp","ginigini","SeaWolf-AI"],"count":6},{"reaction":"👀","users":["Anserwise","quantid","ginigen-ai","SeaWolf-AI"],"count":4},{"reaction":"❤️","users":["Warecube","mayafree","SeaWolf-AI"],"count":3},{"reaction":"👍","users":["mayafree","SeaWolf-AI"],"count":2}],"publishedAt":"2026-09-02T06:16:58.000Z","updatedAt":"2026-09-02T17:01:12.290Z","commentators":[{"_id":"63ac5701c21e60a3e9b58aa7","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/63ac5701c21e60a3e9b58aa7/g6EX7diOpuA94R2ab-rZC.png","fullname":"Dipankar Sarkar","name":"dipankarsarkar","type":"user","isPro":true,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":86,"isUserFollowing":false},{"_id":"6905bc786cb49b1f11d32728","avatarUrl":"https://cdn-avatars.huggingface.co/v1/production/uploads/noauth/504I_H8J9NkiytiHvbx-h.jpeg","fullname":"VIDRAFT_LAB","name":"SeaWolf-AI","type":"user","isPro":false,"isHf":false,"isHfAdmin":false,"isMod":false,"followerCount":305,"isUserFollowing":false}],"url":"/posts/SeaWolf-AI/285774524970283","totalUniqueImpressions":1462,"identifiedLanguage":{"language":"en","probability":0.9242841601371765},"numComments":4}],"numTotalItems":8005}