Image created with gemini-2.5-flash-image with claude-sonnet-4-5. Image prompt: A gleaming chess knight piece stands on an unusual chessboard divided into four quadrants, each displaying different media: musical notes and audio waveforms, flowing text and letters, photographic images, and speech bubbles with sound waves, the knight casting four differently colored shadows in red, blue, green, and amber across the board, dramatic side lighting, rich depth of field, cinematic composition
We’re building real-time speech-to-speech, which will be the default UI between humans and robots F.03 has a 4x more powerful speaker with improved microphone for performance and clarity 👇 https://x.com/adcock_brett/status/1980301303172694209
I am continually surprised about how few applications take advantage of the fact that AI systems can work with video. For example, I can ask Gemini questions about what happens in a video (and not mentioned in a transcript) and get coherent answers including identifying emotion https://x.com/emollick/status/1980695990790418889
Today, we’re one step closer to AI as an operating system. A computer you can talk to, that can see what you see, and take action – all with your permission, all more intuitive than ever. Vision now GA globally + more on today’s @Windows blog: https://x.com/mustafasuleyman/status/1978808627008847997
Open “Ask ChatGPT” and ChatGPT can see the page you’re on to give instant, accurate answers—no tab-switching required. https://x.com/OpenAI/status/1981098271901962439
Yesterday we launched ChatGPT Atlas, our new web browser. In Atlas, ChatGPT agent can get things done for you. We’re excited to see how this feature makes work and day-to-day life more efficient and effective for people. ChatGPT agent is powerful and helpful, and designed to be”” / X https://x.com/cryps1s/status/1981037851279278414
Exclusive: OpenAI to release web browser in challenge to Google Chrome | Reuters https://www.reuters.com/business/media-telecom/openai-release-web-browser-challenge-google-chrome-2025-07-09/
You can also use incognito mode when you don’t want ChatGPT to remember what you are doing in the browser. https://x.com/omarsar0/status/1980688230904144086
Qwen Deep Research just got a major upgrade. ⚡️ It now creates not only the report, but also a live webpage 🌐 and a podcast 🎙️ – Powered by Qwen3-Coder, Qwen-Image, and Qwen3-TTS. Your insights, now visual and audible. ✨ 👉 https://x.com/Alibaba_Qwen/status/1980609551486624237
We’ve built tooling to help you build and ship an e2e document agent in < 5 mins 📄🤖 – and the core is available to everyone! A lot of data is within docs. We already have incredible OCR tooling over docs. The next step is knowledge automation. There’s a lot of agentic https://x.com/jerryjliu0/status/1980759684916408443
Spatial intelligence is so hot right now. And it will only get hotter.”” / X https://x.com/bilawalsidhu/status/1979182318553305597
World models: No one knows what it means, but it’s provocative. It gets the VCs going.”” / X https://x.com/bilawalsidhu/status/1979928032967094730
Alai (@getalai) helps you create high-quality presentations faster. Designed for iteration, quality, and speed, Alai turns rough ideas into decks you’re proud to present. https://x.com/ycombinator/status/1978566353322651683
Massively unexpected update from DeepSeek: a powerful, high-compression MoE OCR model. > In production, DeepSeek-OCR can generate 33 million pages of data per day for LLMs/VLMs using 20 nodes (x8 A100-40G). They want ALL the tokens. You’re welcome to have some too. https://x.com/teortaxesTex/status/1980160624140456370
DeepSeek released an OCR model today. Their motivation is really interesting: they want to use visual modality as an efficient compression medium for textual information, and use this to solve long-context challenges in LLMs. Of course, they are using it to get more training https://x.com/iScienceLuvr/status/1980247935700066468
Text or Pixels? It Takes Half: On the Token Efficiency of Visual Text Inputs in Multimodal LLMs Looks like this paper is also exploring the direction DeepSeek is interested in: representing text more efficiently as images, observing almost half reduction in number of tokens https://x.com/iScienceLuvr/status/1980942325573648703
> by storing the data representation natively as image tiles This must be obvious but just to clarify: DeepSeek does not propose to store *screenshots* of your chat logs. Pixel representation can be ephemeral; what is stored is still tokens, just not *language* tokens. https://x.com/teortaxesTex/status/1980453820632297900
Karpathy is undoubtedly vision pilled. And thanks to this casual DeepSeek drop — so will you.”” / X https://x.com/bilawalsidhu/status/1980598830916939880
DeepSeek https://github.com/deepseek-ai/
@mervenoyann The good perf of DeepSeek models matches with what we observe on PrediBench! https://x.com/AymericRoucher/status/1980196484617523445
Again, I will reiterate this: DeepSeek was literally built by chinas top quant firm, and chinas TOP quants.”” / X https://x.com/hamptonism/status/1980182896049811780
After DeepSeek-V3.2-Exp added TileLang & CUDA ops, many asked: what exactly is TileLang? 🤔 In his post “”TileLang: 80 lines of Python kernel code to reach 95% of FlashMLA’s performance””, developer & Zhihu contributor ryume gives a full breakdown of this new AI programming https://x.com/ZhihuFrontier/status/1980170674112188440
GLM-4.6 providers overview: we are benchmarking API endpoints offered by Baseten, GMI, Parasail, Novita, Deepinfra GLM-4.6 (Reasoning) from @Zai_org is one of the most intelligent open weights models, with intelligence close to GPT-OSS-120b (high), DeepSeek V3.2 Exp (Reasoning) https://x.com/ArtificialAnlys/status/1980777360724226282
For people thinking that DeepSeek-OCR is the first model to render text as images, the University of Copenhagen already did this in 2023 Paper is called “”Language Modelling with Pixels””. They trained a Masked AutoEncoder (MAE) by rendering text as images and masking patches https://x.com/NielsRogge/status/1980559120760791125
We’re seeing a lot of usage around DeepSeek’s new OCR model. Alex packaged it so you can deploy and test it yourself – prompts and sample images included.”” / X https://x.com/basetenco/status/1980924381217104338
DeepSeek-OCR looks impressive, but its core idea is not new. Input “Text” as “Image” — already explored by: LANGUAGE MODELING WITH PIXELS (Phillip et al., ICLR 2023) CLIPPO: Image-and-Language Understanding from Pixels Only (Michael et al. CVPR 2023) Pix2Struct: Screenshot https://x.com/awinyimgprocess/status/1980506449706119642
A more serious thread on the DeepSeek-OCR hype / serious misinterpretation going on. 1. On token reduction via representing text in images, researchers from Cambridge have previously shown that 500x prompt token compression is possible (ACL’25, Li, Su, and Collier). Without https://x.com/Kangwook_Lee/status/1980709454522744902
DeepSeek finally released a new model and paper. And because this DeepSeek-OCR release is a bit different from what everyone expected, and DeepSeek releases are generally a big deal, I wanted to do a brief explainer of what it is all about. In short, they explore how vision https://x.com/rasbt/status/1980642191950090585
I quite like the new DeepSeek-OCR paper. It’s a good OCR model (maybe a bit worse than dots), and yes data collection etc., but anyway it doesn’t matter. The more interesting part for me (esp as a computer vision at heart who is temporarily masquerading as a natural language”” / X https://x.com/karpathy/status/1980397031542989305
🚨 DeepSeek just did something wild. They built an OCR system that compresses long text into vision tokens literally turning paragraphs into pixels. Their model, DeepSeek-OCR, achieves 97% decoding precision at 10× compression and still manages 60% accuracy even at 20×. That https://x.com/godofprompt/status/1980233080213590326
Letsss gooo! DeepSeek just released a 3B OCR model on Hugging Face 🔥 Optimised to be token efficient AND scale ~200K+ pages/day on A100-40G Same arch as DeepSeek VL2 Use it with Transformers, vLLM and more 🤗 https://x.com/reach_vb/status/1980170192392270227
a bunch of OCR models released in past few weeks: ~ deepseek-ocr-3b ~ olmo-ocr-2-7b ~ chandra-ocr-8b ~ nanonets-ocr2-3b ~ paddleocr-vl-0.9B ~ qwen3-vl-dense/moe (general vlm) ~ dots.ocr-3b Will be dropping a detailed comparison soon”” / X https://x.com/HarveenChadha/status/1981055277408669934
NEW DeepSeek OCR model that outperforms dots ocr while prefilling 3x less tokens https://x.com/casper_hansen_/status/1980166248878203093
DeepSeek-OCR has some weird architectural choices for the LLM decoder: DeepSeek3B-MoE-A570M -> uses MHA, no MLA (not even GQA?) -> 2 shared experts (like DeepSeek V2, but V3 only has 1) -> quite low sparsity, activation ratio is 12.5%. For V3 it’s 3.52%, for V2 it’s 5% -> not https://x.com/eliebakouch/status/1980193125202083951
I think Glyph coming out on the same day a) corroborates the results of DeepSeek OCR b) confirms the “they had it lying around for a while” suspicion. Charitably, they learned of Zhipu’s project retracing their steps and sped up the release. Other possibilities are obvious too.”” / X https://x.com/teortaxesTex/status/1980642000006451348
deepseek-ai/DeepSeek-OCR: Contexts Optical Compression https://github.com/deepseek-ai/DeepSeek-OCR
what happened this week with OCR and VLMs? * deepseek-ocr * chandra-ocr * nanonets-ocr2 * paddleocr-vl * qwen3-vl (2B, 32B, Instruct and Thinking) * dots.ocr * olmOCR 2 (based on Qwen2.5-VL) * LightOnOCR (smallies) top 5 trending models on @huggingface are still OCR/VLM! https://x.com/MaziyarPanahi/status/1981421331053760775
DeepSeek-OCR Contexts Optical Compression https://x.com/_akhaliq/status/1980260630780162505
DeepSeek OCR dropped … but honestly, Glyph [1], released the same day, showed something more interesting: 3–4× context compression and infilling cost reduction, no performance hit on long-context QA and summarization, which is much less trivial than OCR in many cases. If that https://x.com/arankomatsuzaki/status/1980722682246398069
🚀 DeepSeek-OCR — the new frontier of OCR from @deepseek_ai , exploring optical context compression for LLMs, is running blazingly fast on vLLM ⚡ (~2500 tokens/s on A100-40G) — powered by vllm==0.8.5 for day-0 model support. 🧠 Compresses visual contexts up to 20× while keeping https://x.com/vllm_project/status/1980235518706401405
Google prepares Genie 3 public experiment with AI worlds https://www.testingcatalog.com/google-prepares-genie-3-public-experiment-with-ai-generated-worlds/
We’re releasing the full FinePdfs source code — plus new datasets and models! 🚀 📚 Datasets: • OCR-Annotations — 1.6k PDFs labeled for OCR need • Gemma-LID-Annotation — 20k samples per language (annotated with Gemma3-27B) 🤖 Models: • XGB-OCR — OCR classifier for PDFs https://x.com/HKydlicek/status/1980319822585143498
📣 Announcing MUSI: 1st Multimodal Spatial Intelligence Workshop @ICCVConference! 🎙️All-star keynotes: @sainingxie, @ManlingLi_, @RanjayKrishna, @yuewang314, and @QianqianWang5 – plus a panel on the future of the field! 🗓 Oct 20, 1pm-5:30pm HST 🔗 https://x.com/songyoupeng/status/1975811164765643058
Whale insanity. «Theoretically unlimited context architectures» through turning old text into «vision» tokens at 10x and higher ratios. How crazy is that? But actually I think… we need to go even further beyond. Fully multimodal encoders. Language for intelligent machines. https://x.com/teortaxesTex/status/1980165682516869575
📢 The MUSI workshop is happening TOMORROW (Oct 20) between 1 pm-5:30 pm! 🙋♀️🙋Feel free to add any (spicy) questions on multimodal spatial intelligence you want me to ask during the panel discussion at https://x.com/songyoupeng/status/1980110329473593524
Building State-of-the-Art VLAs with Zero Modification. [📍 Save paper + code] Everyone’s trying to design the perfect Vision-Language-Action model. Custom heads. Special action tokens. Extra modules. But what if the best architecture needs none of that? VLA-0 turns a https://x.com/IlirAliu_/status/1979086524672098693
want to fine-tune models for OCR/document??? 📑 two tutorials for you 🫡 > fine-tune Kosmos2.5 with grounding: if you have data with bounding boxes + text inside > fine-tune Florence-2 on DocVQA: if you search for answers in a document plug and play with other VLMs! 💗 https://x.com/mervenoyann/status/1981657235785728010
Wow OCR models are taking off in vLLM 😍 Small but powerful 💪 Enjoy this fast OCR model from @staghado ✌️”” / X https://x.com/vllm_project/status/1981579850436751611
Moondream 3 can parse complex parking signs in one step. Prompt: “”extract sign details”” → JSON of each rule + transcription. No OCR stack, no regex: just vision that understands structure. ⚡️Fast, cheap, grounded vision AI. https://x.com/moondreamai/status/1980405287531254089
We’re updating olmOCR, our model for turning PDFs & scans into clean text with support for tables, equations, handwriting, & more. olmOCR 2 uses synthetic data + unit tests as verifiable rewards to reach state-of-the-art performance on challenging documents. 🧵 https://x.com/allen_ai/status/1981029159267659821
There’s been a crazy OCR mania for the last couple of days 👀 And you can 1-click deploy most of these models directly from the Inference Endpoints catalog 🔥 https://x.com/ErikKaum/status/1981750508982268330
one of the big motivations behind olmOCR 2’s use of RLVR with binary unit tests. the ability to easily define unit tests for model failures + retrain makes iteration really easy tech report out 👉 https://x.com/kylelostat/status/1981380820658180310
You might have seen a lot of OCR release recently… Here is another one, introducing 🦉 LightOnOCR-1B A fully end-to-end differentiable VLM model competing with all the latest releases while being much faster🚀 https://x.com/staghado/status/1981379888301867299
olmOCR – Open-Source OCR for Accurate Document Conversion https://olmocr.allen.ai/blog
Deploy your favorite OCR models with few-clicks directly from Hugging Face 🔥 📷we’ve added the latest bleeding edge OCR models to the Inference Endpoints catalog to make it easy for you to get started! links 👇 https://x.com/ErikKaum/status/1980965155145216336
there’s a new OlmOCR model that outperforms other OCR models, with Apache 2.0 license 🔥 and it costs only $178 to parse million pages 🤯”” / X https://x.com/mervenoyann/status/1981040748133826918
I’m excited to announce that Chandra OCR is open source! – Full layout information – Extracts and captions images and diagrams – Strong handwriting, form, table support – Works with transformers and vLLM https://x.com/VikParuchuri/status/1980667137606971423
Hugging Face just unveiled FineVision: The largest & cleanest open dataset for VLMs A meticulously curated corpus of 24 million samples, unifying 200+ sources into 185 subsets via a semi-automated, human-in-the-loop pipeline. Outperforms existing open mixtures, accelerating https://x.com/HuggingPapers/status/1981093262912819418
Introducing our new tiny vision language model: LFM2-VL-3B 👀 > Expanded multilingual visual understanding: English, Japanese, French, Spanish, German, Italian, Portuguese, Arabic, Chinese, Korean > 51.8% on MM-IFEval (instruction following) > 71.4% on RealWorldQA (real-world https://x.com/LiquidAI_/status/1980985540196393211
I don’t use a Mac, so have no thoughts on the new browser.🤷♂️”” / X https://x.com/emollick/status/1980698932029321406
Unseeable prompt injections in screenshots: more vulnerabilities in Comet and other AI browsers | Brave https://brave.com/blog/unseeable-prompt-injections/
Sora 2 Pro is very impressive in my tests but really needs its own interface – using a storyboard maker shoved inside a video creation tool shoved inside of a drafts folder of a social media app is awkward.”” / X https://x.com/emollick/status/1980132516406473126
Introducing: Interactive Sora! A choose-your-own adventure GAME powered by Sora 2. Every choice spins up a brand-new scene instantly. Open source, link here: https://x.com/mattshumer_/status/1978848940083839162
we just updated the model comparison on our blog for you 🫡 added Chandra, OlmOCR-2, Qwen3-VL and their averaged OlmOCR score! https://x.com/mervenoyann/status/1981396054634615280
Introducing Tahoe-x1 (Tx1) by @tahoe_ai. A 3-billion-parameter, single-cell foundation model that learns unified representations of genes, cells, and drugs, achieving state-of-the-art performance across cancer-relevant cell biology benchmarks, open-sourced on @huggingface. 🧵 https://x.com/nalidoust/status/1981760790551298524
MEG-GPT: A transformer-based foundation model for magnetoencephalography data MEG is another noninvasive neuroimaging technique similar to fMRI. Don’t think I’ve seen a large-scale foundation model for this type of data yet, very interesting! “”we introduce MEG-GPT, a https://x.com/iScienceLuvr/status/1980945270369399234
World-in-World: World Models in a Closed-Loop World https://world-in-world.github.io/
Building an infinitely learning geolocation model with online RL – sdan blog https://blog.sdan.io/geospot-infinity/
Qwen just released Qwen3-VL on Hugging Face The most powerful vision-language model in the Qwen series, with comprehensive upgrades across text understanding, visual reasoning, and long context video analysis. From GUI operations to 1M context. https://x.com/HuggingPapers/status/1980809413045940553
