Liquid AI (@liquidai) header
Liquid AI (@liquidai) profile picture

Build efficient general-purpose AI at every scale.

Cambridge, MA
Joined March 2023
Photo by Liquid AI (@liquidai)Photo by Liquid AI (@liquidai)Photo by Liquid AI (@liquidai)Photo by Liquid AI (@liquidai)Photo by Liquid AI (@liquidai)Photo by Liquid AI (@liquidai)Photo by Liquid AI (@liquidai)Photo by Liquid AI (@liquidai)Photo by Liquid AI (@liquidai)Photo by Liquid AI (@liquidai)
Introducing 𝙻𝙵𝙼𝟹: our new generation of Liquid Foundation Models. With a new tri-hybrid architecture and new multimodal encoders, the LFM3 family is the ideal model suite for specialized intelligence. > 6 model instances, across text, audio, vision, and omni modalities, generative and decision models. > 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱: text, 256k context window. > 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱: vision-language model with video-input support for spatial understanding > 𝙻𝙵𝙼𝟹-𝙰𝚞𝚍𝚒𝚘-𝟸𝙱: tool calls, voice chat, reasoning in 18 languages. > 𝙻𝙵𝙼𝟹-𝙾𝚖𝚗𝚒-𝟸𝙱: an omni model combines vision, audio, and language. > 𝚍𝟷-𝟸𝙱 and 𝚍𝟷-𝙾𝚖𝚗𝚒-𝟸𝙱: decision models with typed outputs and calibrated probabilities in a single pass. Omni adds audio input. > fine-tune on 1 GPU. > a new tri-hybrid architecture > architecture optimized for 14 processor classes across GPUs, CPUs, and NPUs. > 𝙻𝙵𝙼𝟹-𝚅𝙻-𝟸𝙱 is a beast. It performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters. > 𝙻𝙵𝙼𝟹-𝟷.𝟽𝙱 is #1 on the Pareto frontier of device intelligence 𝚙𝚒𝚙𝚎𝚝𝚝𝚎.𝚕𝚒𝚚𝚞𝚒𝚍.𝚊𝚒 Every modality for specialized intelligence. Pick the one that fits your use case, customize it, and run it on-device. 1/ 🧵
539
1,255
738
9,060
2,390,102
Liquid AI retweeted
time to update your decision model.
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
1
1
5
285
Liquid AI retweeted
Super proud of the traction d1 got since its release. Now with audio and many new primitives, we're ready to shape the next stage of decision models 🚀
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
7
2
35
1,393
Liquid AI retweeted
Liquid AIがd1.2とd1.2-ultraをリリース!! 本家Jevとか相手にならないレベルじゃん… 出すモデル全てイケてる、Liquid AI 一旦、最強の意思決定モデルはd1.2-ultraってことか… d1.2はテキスト・画像・音声を扱える最先端のオムニモーダル意思決定モデルで、Artificial Analysis Decision Index v2.1で1位(75点)を獲得。 音声も新たに追加。 新しく追加された主な機能: ・請求書や領収書からの値読み取り(金額・日付など。該当なしは「not found」と返す) ・テキスト内の特定情報(PIIなど)のスパン抽出+位置情報 ・画像内の物体位置特定(バウンディングボックス。複数対応で個数も) ・数値の推定(最良値+範囲を返す) ※これアツい🔥 ・複数選択肢の適用判定(各オプションに確率付き) Decision Indexは正確性だけでなく「キャリブレーション」も重視したスコア。 正しい答えで自信が高く、間違った答えで過信しないほど高得点になる。 d1.2-ultraのキャリブレーション誤差は0.8と非常に低く、信頼度閾値を設定して「高信頼は自動処理、低信頼は人間レビュー」といった運用がしやすい。 現状SOTA急Decisionモデルだから、多分APIのみで重みは出ない気がするけど 入力トークンのみ課金(d1.2が$0.02、ultraが$0.042 / 100万トークン)でコスト効率も良く、ドキュメント処理・チケットルーティング・視覚検査・音声ベースの判定などに向いている。
Replying to @liquidai
d1.2-ultra ranks first among decision models on the @ArtificialAnalysis Decision Index v2.1, and for less than half its price, d1.2 offers the best trade-off between quality and cost. Both models are available today on the Liquid API, at $0.02 for d1.2 and $0.042 per million input tokens for d1.2-ultra. /2
4
28
2,227
Liquid AI retweeted
A new refresh of our d1.1 model, now with audio input:
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
1
2
11
501
Liquid AI retweeted
🤯决策模型被颠覆了!Liquid AI 新出的 d1.2-ultra 直接拿下决策榜第一名!分数高达 75 分,把一堆大厂模型远远甩在后面!d1.2 更是性价比杀手! 更狠的是,它现在能同时看文字、看图片、听声音,然后直接告诉你答案和靠谱程度,不用再绕一大圈生成废话。 举几个例子就懂了: 🔹扔一张发票进去,它能准确读出金额 🔹丢一张照片,它能帮你框出东西在哪 🔹让它找合同里的人名、电话,它能精准标出来 🔹问照片里有多少人,它还会给你一个范围,不瞎猜 概率也特别准,说80%就真的接近80%,不再是那种“我很自信但其实在瞎扯”的模型。 而且价格低到离谱:普通版只要 2分钱 就能处理一百万字的输入。 需要做判断、提取信息、看图分类的任务,直接上它就行了,又准又便宜。 #LiquidAI #d12 #决策模型 #AI新模型
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
10
5
38
3,109
wake up, randomly a new SOTA model from Liquid appears, many such cases in recent weeks 🫡
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
3
3
13
704
d1.2-ultra is the #1 model on the Artificial Analysis Decision Index v2.1, with d1.2 placing #6 @liquidai has released d1.2-ultra and d1.2, updating their d1-ultra and d1.1 flagship decision models. Key results: ➤ d1.2-ultra is the #1 model on AA-Paperclip. With a score of 76, d1.2-ultra is the leading model on AA-Paperclip, our first-party private benchmark that assesses decision-making and extraction capabilities. It leads the next-best model, Drex Max v1.1, by 3 points ➤ d1.2-ultra and d1.2 improve over their predecessors. d1.2-ultra scores 75, a 4-point gain over its predecessor d1-ultra, while d1.2 scores 69, a 3-point improvement over d1.1 ➤ d1.2-ultra and d1.2 both sit on the Capability vs. Cost Pareto frontier. d1.2-ultra and d1.2 cost $133 and $63, respectively, per one million Artificial Analysis Decision Index tasks and both models sit on the Capability vs. Cost Pareto frontier Additional model details: ➤ Context window: 64K tokens ➤ Pricing: $0.042 and $0.02 per 1M input tokens for d1.2-ultra and d1.2, respectively ➤ Multimodality: Text and image input ➤ Providers: Available through Liquid AI’s first-party API
30
26
7
270
25,950
Liquid AI retweeted
Liquid AI released d1.2 and d1.2-ultra, decision models that now support audio input and five new task types. They can extract values from documents, find every relevant passage, locate objects in images, estimate quantities, and select all applicable options. Unlike generative models, they return structured decisions with calibrated probabilities.
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
1
5
491
d1 got even better! @liquidai just released d1.2, the most powerful decision model yet. Huge fan of LiquidAI models, congrats on the release!
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
2
10
721
Liquid AI retweeted
Start building today !...
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
1
3
498
Liquid AI retweeted
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
1
6
67
2,764
Liquid AI retweeted
liquid's new decision models report how sure they are... liquid released d1.2 and d1.2-ultra. they take text, images and audio. the interesting bit is calibration. when the model says 80%, it should be right about 80% of the time, so you can set a threshold and send low-confidence cases to a human. new tasks in the api: - read a value from an invoice, with the characters it came from, and say so if the invoice doesn't show it - find every passage of a kind in a text, or every box for an object in a picture - pick every option that applies, or estimate a number with a range caveat: the #1 spot on the artificial analysis decision index (75) is liquid's own claim, so treat it as a vendor number until you test it. takeaway: run it on your own messy documents and see where the confidence threshold lands. via @liquidai
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
1
3
394
Liquid AI retweeted
I love this direction for decision models! Omni input and rich output forms. Many complex legacy ML systems can be replaced with such models and in new systems a lot of new capabilities can be added with few lines of code. Decision models are like swiss army knife.
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
1
2
383
Liquid AI retweeted
A span, a box, a range, a set of choices — d1.2 answers with types instead of prose, so nothing downstream has to regex JSON out of a chat reply, and the "number with a range" output hands the next step an uncertainty band to threshold on. That design also explains the price shape: extraction-shaped calls fit smaller models, and an agent can save the expensive planner for the one step that needs it. First place at 75 on the Artificial Analysis Decision Index comes from Liquid's own screenshot, and that index rewards exactly the output modes this model built for, so read the ranking as a fit rather than a general win. Nothing in the post shows behaviour when the audio runs long or the image holds a crowd of objects, which matters more here than for a chat model, since one dropped detection silently returns an empty list.
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
2
4
416
Liquid AI retweeted
An invoice tool should be allowed to say "not found." Liquid AI's d1.2 and d1.2-ultra add structured extraction to its decision-model API. Its announcement describes reading values, returning text spans and locating objects in images, alongside audio input. My first pilot would be boring: supplier, total and due date from non-confidential sample invoices. Include missing totals, unclear scans and conflicting dates. Count wrong answers and missed fields separately. Build a review queue, not "AI pays every invoice." Show each value beside its source text, send uncertain cases to a person, and keep payment approval separate. Liquid says there is no public benchmark for these new tasks yet. Its internal results are not proof of accuracy on your invoices. Check confidence thresholds on your own examples too. The input-to-output diagram below is Liquid's, not my test.
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
1
2
312
Liquid AI retweeted
New decision models d1.2 and d1.2-ultra which accept text, images and audio!
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
1
3
11
575
Liquid AI retweeted
Wow! Liquid AI just released the best omnimodal AI decision maker! This is huge!
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
3
11
117
7,622
Liquid AI retweeted
Big release today: d1.2-ultra and d1.2 are now live on the Liquid API ⚡️ They’re of course stronger than our previous iterations and come with super useful new capabilities: > Same typed answers from our open d1-2B and d1-Omni-2B: extract a number, a date or a name out of a document with the exact place it came from, bounding boxes around what you describe in a photo, estimates with a range, and select several options at once. > Audio 🎧: send a clip and ask about it, or just ask your question out loud!
Today, we release d1.2 and d1.2-ultra, the state-of-the-art omnimodal decision models. d1.2 and d1.2-ultra both can process text, images, and audio with new output capabilities. > #1 on the @ArtificialAnlys Decision Index (75) > Read a value from the input (text, number or date) > Find every passage of a kind in a text (spans) > Locate things in a picture (box for one, boxes for every instance) > Estimate a number (number with a range) > Pick every option that applies (choice) 1/ 🧵
8
5
42
1,495
👀 – at Highlands-Baywood Park, CA
d1.2-ultra is the #1 model on the Artificial Analysis Decision Index v2.1, with d1.2 placing #6 @liquidai has released d1.2-ultra and d1.2, updating their d1-ultra and d1.1 flagship decision models. Key results: ➤ d1.2-ultra is the #1 model on AA-Paperclip. With a score of 76, d1.2-ultra is the leading model on AA-Paperclip, our first-party private benchmark that assesses decision-making and extraction capabilities. It leads the next-best model, Drex Max v1.1, by 3 points ➤ d1.2-ultra and d1.2 improve over their predecessors. d1.2-ultra scores 75, a 4-point gain over its predecessor d1-ultra, while d1.2 scores 69, a 3-point improvement over d1.1 ➤ d1.2-ultra and d1.2 both sit on the Capability vs. Cost Pareto frontier. d1.2-ultra and d1.2 cost $133 and $63, respectively, per one million Artificial Analysis Decision Index tasks and both models sit on the Capability vs. Cost Pareto frontier Additional model details: ➤ Context window: 64K tokens ➤ Pricing: $0.042 and $0.02 per 1M input tokens for d1.2-ultra and d1.2, respectively ➤ Multimodality: Text and image input ➤ Providers: Available through Liquid AI’s first-party API
7
2
114
3,633
Liquid AI retweeted
Our new Liquid AI d1.2-ultra scores 75 on the @ArtificialAnlys Decision Index (#1). Our new d1.2 (non-ultra) is not far behind at less than half the per-token price of the ultra.
1
3
16
994