Recent posts
👍️
We created a pitch deck to tell a handful of VC firms about us and what we were up to (a fun experience!). Here’s a few slides about our background and some of the things we’ve worked on from the pitch deck (it was fun putting together the list of people in our teams who have gone on to found a whole range of exciting companies). We are delighted to have selected @radicalvcfund and @khoslaventures to lead our initial funding round, along with participation from @lightspeedvp, @kleinerperkins, Doerr Capital (@johndoerr), and Alphabet (@Google). We’ll be working with them to close our seed round over the next few weeks.
オフィスで一番着てる人の多いTシャツは何故か確実にModalのTシャツ お世話になっております 🙏
Congrats to Sakana AI on shipping Namazu! Happy to power Namazu's ~1T-param model for live web search + code execution on Modal.
Kimi K2.6ベースの新バージョンのモデルをAPIで先に公開です。Sakana Chatも今後アップデート予定。
🐟 Sakana Namazu API 公開 🐟 本日、Sakana AIは大規模言語モデル「Namazu」をアップデートし、API「Sakana Namazu（サカナ・ナマズ）」として提供を開始しました。 Sakana Namazu API: https://sakana.ai/namazu 🐟
帰宅。 #ICFPC2026 お疲れ様でした。個人的にはKaggleのNeuroGolf 2026やNeurIPS 2025 Code Golfに通じるところを感じて、それらを参考にしつつCodex + {GPT5.6 Sol, Fugu-Ultra v1.1} をあれこれ工夫しながら叩いてました。結果は振るわず反省も多いが、とりあえず楽しかった！運営の方々には感謝🙏
今年もICFPC2026のチームUnagi（@iwiwi, @sulume, @wata_orz, @toslunar, @chokudai）で参加しました。リポジトリとビジュアライザも公開しておきました！ https://github.com/icfpc-unagi/icfpc2026 https://icfpc-unagi.github.io/icfpc2026/
出ます！
今日から72時間コンテストのICFPC！いつもの6人チームで出るよ！2013年から出ていて6割くらい1位取ってると思うけど、今年はAI強いしどうかなー？がんばるよー！ チームのページ：https://icfpc-unagi.github.io/
What competing in the ICFP Programming Contest looks like in 2026
こちらが今年のICFPCのコーディング環境となります（XREAL One Pro + mutalk 2）
新刊『検索システム』を著者の佐藤竜馬先生 @joisino_ よりご恵贈頂きました！検索の基本からベクトル検索、RAG、LLMによる生成検索まで幅広く扱っていて良さそうです。ありがとうございます！🙏 https://www.amazon.co.jp/dp/4065429714
#ICML2026 でと発表したUnMaskForkについてのブログを出しました。複数の拡散言語モデルを協調させ推論時スケーリングする手法です。 @takkyuO2 との共同研究です。 この論文が出来るまでの過程はなかなか面白かったです。まず興味深い発見として、DreamCoder等のMDLMでは定番の「温度による多様性」 がほぼ使えませんでした。温度を0よりぐっと上げたり、デコード手法を触って確率性を入れようとすると、品質が急落するんですよね。そこで、複数の拡散言語モデルを混ぜ合わせるという少し奇妙な方法を試したところ、これが驚くほどうまく機能したという。 以前発表したAB-MCTS (NeurIPS'25)に続く論文となり「複数モデルの協調＋推論時スケーリング」シリーズが作れたのも嬉しいです。
Can test-time scaling work for diffusion language models? In our #ICML2026 paper "UnMaskFork," we show that having multiple masked diffusion language models collaborate on a single answer improves performance on coding and math tasks. Blog: https://t.co/FZ25e6XCws Test-time scaling is an actively researched technique that boosts LLM performance by using inference-time compute, for example, by having a model think longer or repeatedly refine its answers. This allows us to enhance performance simply by increasing computation during inference without relying on additional training, giving us the flexibility to balance compute costs and performance based on the specific use case. Unlike standard LLMs that generate text left-to-right, masked diffusion language models (MDLMs) generate text by gradually filling in a fully masked sequence. MDLMs can generate multiple parts of a sequence in parallel, offering potential speed-ups, and they can generate flexibly while seeing the entire sequence at once. This makes them an actively studied new paradigm in language modeling. We found that the standard LLM approach of "raising the temperature to increase randomness and generate diverse answers" does not work well for MDLMs like Dream-Coder. Instead of relying on this randomness, our proposed method, UnMaskFork (UMF), creates diversity through "model switching." Multiple MDLMs share the task of unmasking a single answer, and we use Monte Carlo Tree Search to search for a promising sequence in which different models handle different stages. Each model picks up where the others left off, filling in the parts it is most confident about. This collaborative approach allows us to explore diverse answers while maintaining generation quality, consistently outperforming existing test-time scaling methods on coding benchmarks and scaling effectively on math as well. Test-time scaling is also crucial for advancing MDLMs, and our work shows that UMF can sidestep the difficulties specific to them. UMF requires no additional training or changes to the models; it works simply by combining pre-trained models at inference time. This allows us to leverage the diversity of diffusion language models trained on different data and with different methods to improve performance. We believe the value of UMF will only grow as more diverse MDLMs emerge. This work is part of our broader research into "collective intelligence of AI," alongside methods like AB-MCTS and Sakana Fugu that have multiple LLMs collaborate. We'll continue pursuing research that turns model diversity into a source of strength. For details of the algorithm and illustrative examples showing how this collaboration works, please see our blog and paper. Paper: https://t.co/4JC9SYdTyX 🐟
参加してます
【ライブ】フィジカルAI政策に関する対外発信イベント 経産省｜NEDO、NVIDIA、Noetra【LIVE】(2026年7月16日) ANN... https://www.youtube.com/live/n8g5eWD18Xo?si=6sHoRKZMmxvGNIfO
お世話になっております 🙏
【ニュースリリース】 さくらインターネットが提供するベアメタル型GPUクラウドサービス「高火力 PHY」は、Sakana AI株式会社が「GENIAC」第4期で取り組むプロジェクトの計算基盤として採用されました。 https://www.sakura.ad.jp/corporate/information/newsreleases/2026/07/16/1968225411/
CSOはここではChief Strategy Officerの略らしいです
数週間分の戦略リサーチを、数時間で。 あなたのVirtual CSOとして働く、Sakana Marlin。 最初のテーマを渡してみる： https://sakana.ai/marlin 🐟
全問正解は流石にえぐい ICFPCにも出てくれ、勝負や @OpenAI ！！
[16:58] OpenAI が最後の E 問題を正解し、全問正解となりました。 制限時間の 7 時間のうち 6 時間を消費するギリギリの戦いでした。2026 年 7 月 9 日は、競技プログラミングにおける人類の叡智の限界を、わずかにではありますが明確に上回った、歴史的な日となりました。
We’re at Hall A #1701 right now! Come discuss how we ensemble diffusion LMs for test-time scaling. #ICML2026
"UnMaskFork: Test-Time Scaling for Masked Diffusion via Deterministic Action Branching" will be presented at #ICML2026 Paper: https://t.co/4JC9SYdTyX We introduce UnMaskFork, a test-time scaling framework for Masked Diffusion Language Models (MDLMs). Using Monte Carlo Tree Search, it explores diverse generation paths by dynamically switching between multiple pre-trained MDLMs to collaboratively generate the text. Evaluations on coding and mathematical reasoning tasks show that UnMaskFork consistently outperforms standard Best-of-N and other tree search baselines. The results demonstrate that deriving search space diversity from multiple distinct models is a highly effective test-time scaling strategy for MDLMs.
Come say hi and let’s chat about how our tool can help you better understand your pretraining data! Hall A #900 #ICML2026
Presenting a poster at #ICML2026 soon! SoftMatcha 2: A tool that searches a TRILLION-token corpus in ~0.1s ⚡️. Semantic substitution, insertion, and deletion supported. July 8, 10:30 AM––12:15 PM · Hall A, #900 https://icml.cc/virtual/2026/poster/60653 https://softmatcha.github.io/v2/
Just landed in Seoul 🇰🇷 for #ICML2026! I'll be presenting three papers this week, so stop by if you're around! 🙌 SoftMatcha 2: A Fast and Soft Pattern Matcher for Trillion-Scale Corpora 🗓️ Wed Jul 8, 10:30 AM - 12:15 PM KST 📍Hall A #900 https://t.co/LhrjVtMFO6 https://t.co/HNxk0aK5WS https://t.co/hVDXc60F5C UnMaskFork: Test-Time Scaling for Masked Diffusion via Deterministic Action Branching 🗓️ Wed Jul 8, 2:30 PM - 4:15 PM KST 📍Hall A #1701 https://t.co/QCWAMl1MDd https://t.co/xvHuPsXpQR Feedback-to-Rubrics: Can We Extract Expert Criteria from Inline Comments? 🗓️ Sat July 11, 3:45 PM - 5:00 PM KST 📍Room 401, Workshop on Human-AI Co-Creativity: Advances, Opportunities, and Challenges https://t.co/6FurSvynlO https://t.co/jIruNmCcvZ
Sakana AI is heading to #ICML2026 in Seoul (July 6–11)! 🐟🇰🇷 Our team will present 11 papers spanning multi-agent coordination, sparse and efficient LLMs, test-time scaling, long-term memory, and agent benchmarks. A thread of everything we're presenting:
