Discussion

• @smcgrath.phd Scott McGrath on bluesky

  Google has released Gemini 2.5 Deep Think, its first public multi-agent reasoning model.  The system explores multiple ideas in parallel to solve complex problems, achieving state-of-the-art results on coding and reasoning benchmarks.  —  #MLSky

• @timkellogg.me Tim Kellogg on bluesky

  Gemini 2.5 Deep Think is here  —  “It is a variation of the model that recently achieved the gold-medal standard at this year's International Mathematical Olympiad (IMO)”  —  Apparently it's good at creative writing as well  —  blog.google/products/gem...  [images]

• @officiallogank Logan Kilpatrick on x

  Quick correction: this is a variation of our IMO gold model that is faster and more optimized for daily use! We are also giving the IMO gold full model to a set of mathematicians to test the value of the full capabilities. Too many new and powerful models 😅

• @googledeepmind @googledeepmind on x

  For researchers, scientists, and academics tackling hard problems: Gemini 2.5 Deep Think is here. 🤯 It doesn't just answer, it brainstorms using parallel thinking and reinforcement learning techniques. We put it into the hands of mathematicians who explored what it can do ↓ [vide…

• @altryne Alex Volkov on x

  Had a chance to test run deep think for a while, and it's dope! Honestly the advances in AI lately are mind boggling, and to think that “reasoning” and test time compute was not even a thing a year ago!!

• @emollick Ethan Mollick on x

  Had early access to Gemini with Deep Think. Very good model, big gains over standard Gemini 2.5 Pro for a lot of problems. Here is the first attempt at the starship control panel prompt I try with every model. First time I have seen a model make a 3D interface in response. [video…

• @_philschmid Philipp Schmid on x

  Gemini Deep Think is now available for all Ultra users! Gemini 2.5 Deep think is a enhanced reasoning mode to explore and combine multiple solution paths. - Parallel thinking to explore and combine multiple solution paths. - State-of-the-art performance on LiveCodeBench V6 and [i…

• @jeffdean Jeff Dean on x

  See mathematician Michel van Garrel talking about how our latest Gemini Deep Think model was able to prove a conjecture using a very different approach than he was considering.

• @vibedonnie Donnie on x

  Google has begun rolling out it's brand new Gemini 2.5 Deep Think feature to Gemini Ultra subscribers today [video]

• @mariusfanu Marius Fanu on x

  Google DeepMind is rolling out Gemini 2.5 Deep Think, a multi-agent reasoning model. For $250-per-month Ultra subscribers, it arrives in the Gemini app today. The system spawns several agents to attack a problem in parallel, boosting accuracy but raising compute costs. Google [im…

• @simonw Simon Willison on x

  Impressive that this is the same model they used for the International Mathematical Olympiad a few weeks ago

• @cedric_chee Cedric on x

  Gemini 2.5 Deep Think is rolling out in the Gemini app for Google AI Ultra subscribers. It can improve both the aesthetics + functionality of web dev tasks. Impressive! And SoTA performance across LiveCodeBench V6 (86.6%) and HLE (34.8% without external tools) [image]

• @officiallogank Logan Kilpatrick on x

  Gemini Deep Think, our SOTA model with parallel thinking that won the IMO Gold Medal 🥇, is now available in the Gemini App for Ultra subscribers!! Should we put it in the Gemini API next? [image]

• @giacomomiolo Giacomo Miolo on x

  gemini 2.5 deep think > 1M context window > 192k token output (incredibly long) > approaches human performance on “cover your tracks” and “strategic rule breaking” [image]

• @nanulled Nano on x

  Gemini 2.5 Deep Think Model Card: it's not superhuman but similar to gold IMO model & “approaches human level” on stealth evals more interested in learning “novel rl techniques that can leverage more multi-step reasoning,” (candidate: MARL with verification/voting for each step) …

• @hangsiin @hangsiin on x

  ‘Gemini 2.5 Deep Think IMO 2025 results are computed as pass@1 while all the other results coming from https://matharena.ai/ are best of 32.’ [image]

• @sundarpichai Sundar Pichai on x

  We're bringing a version of Deep Think that achieved gold-medal status at IMO to Ultra subscribers in the @Geminiapp (+ the official version is now in the hands of mathematicians). Toggle it on when reasoning through complex scientific literature, tackling a coding problem that […

• @jeffdean Jeff Dean on x

  Starting today, we're offering Deep Think in the Gemini app for Google AI Ultra subscribers, and we're giving a collection of mathematicians access to the full version of the Gemini 2.5 Deep Think model that achieved gold medal🏅 level performance in the recent IMO competition. [i…

• @koltregaskes @koltregaskes on x

  Google DeepMind launches Gemini 2.5 Deep Think, brainstorms multiple solutions in parallel using reinforcement learning. Outperforms rivals on benchmarks like LiveCodeBench V6 (87.6%) and Humanity's Last Exam (34.8%) without tools, and achieving bronze on IMO 2025 with an [image]

• @sarah_a_bentley Sarah Bentley on x

  Gemini 2.5 Deep Think benchmarks. [image]

• @engineerrprompt @engineerrprompt on x

  Gemini 2.5 Deep Think from @GoogleDeepMind is here! This model is different, its built for the hard problems! A version of this won Gold on IMO 2025. Its simply amazing that you can use model of this capability in @GeminiApp. Its probably an overkill for most tasks, so use [video…

• @aistudio30 @aistudio30 on x

  Gemini 2.5 pro Deep Think Benchmark Holy shit... [image]

• @jonathancochran @jonathancochran on x

  The race isn't about who builds the smartest model. It's about who can turn “parallel thinking” into real breakthroughs in health, energy, and beyond. Is Gemini 2.5 Deep Think a step closer?

• @andrewcurran_ Andrew Curran on x

  Reposting as the previous thread broke. Gemini 2.5 Deep Think is live today for Ultra subscribers. This is a variation of the model that won the gold-medal at this year's International Mathematical Olympiad. [image]

• @daniel_mac8 Dan Mac on x

  🚨 BREAKING: Google is launching Gemini 2.5 DeepThink to Gemini ultra subscribers 🥇 it's a variant of the model that won IMO gold All the talk about GPT-5 coming and Deepmind snuck this one in, not to be upstaged Benchmarks look incredible. Have me considering an Ultra sub [image]

• @legit_api @legit_api on x

  wow - Gemini 2.5 Deep Think [image]

• @kimmonismus @kimmonismus on x

  Gemini Deep Think in the Gemini app for Google AI Ultra subscribers released What a day! Gemini DeepThink could change the entire playing field. Now it will be exciting to see if OpenAI will respond today! [image]

• @harshithlucky3 @harshithlucky3 on x

  Gemini 2.5's new Deep Think is now available for Google AI Ultra subscribers > It uses “extended, parallel thinking” to solve problems better and come up with creative ideas An even smarter version is going to a small group of experts for research

• @sam_witteveen Sam Witteveen on x

  Gemini Deep Think is here and it is a very different kind of model with some impressive skills! Check out the full video in the reply. [image]

• @demishassabis Demis Hassabis on x

  Gemini 2.5 Deep Think now available for Ultra subscribers! Great at tackling problems that require creativity & planning, it finds the best answer by considering, revising & combining many ideas at once. A faster variation of the model that just achieved IMO gold-level. Enjoy!

• @ai_for_success AshutoshShrivastava on x

  Breaking: Google DeepMind has released Gemini 2.5 Deep Think for Ultra users. Incredible benchmarks: HLE - 34.8% without external tools Live Code Bench - 86.6% AIME 2025 - 99.2% Did Google DeepMind just solve math? Whatever the case, Google delivers with no hype 🔥 Details 👇 [imag…