home.social

#o3 — Public Fediverse posts

Live and recent posts from across the Fediverse tagged #o3, aggregated by home.social.

  1. OpenAI’s o3: The Reasoning Engine Redefining AI for Coders, Scientists and Enterprises OpenAI's o3 model, released April 2025, masters visual reasoning, tool use, and tough benchmarks in code...

    #SupplyChainPro #AI #Agents #AIME #math #ChatGPT #tools #o4-mini #OpenAI #o3 #reasoning

    Origin | Interest | Match
  2. New series I'm trying on the channel, let me know what you think. Short Indie game first impression style videos. #indiegame #o3 #gaming

    youtu.be/hJwiadpoZVw

  3. Freaky Friday auf Ö3 - Heute Lieder mit unsinnigen Texten.
    How much ist the fish hatten wir schon. 😂
    #radio #ö3 #freaky friday

  4. Freaky Friday auf Ö3! Heute mit Begrüßungs-Songs. 😊
    #Ö3 #freakyfriday #radio

  5. Freaky Friday auf Ö3! Heute große Fußballsongs und Stadionhymnen.
    #Ö3 #Radio #FreakyFriday

  6. 🧠 Il nuovo modello #Gemini Flash mostra la stessa precisione di #o3 in attività agentiche di utilizzo del browser. 

    👉 I dettagli: linkedin.com/posts/alessiopoma

    ___
    ✉️ 𝗦𝗲 𝘃𝘂𝗼𝗶 𝗿𝗶𝗺𝗮𝗻𝗲𝗿𝗲 𝗮𝗴𝗴𝗶𝗼𝗿𝗻𝗮𝘁𝗼/𝗮 𝘀𝘂 𝗾𝘂𝗲𝘀𝘁𝗲 𝘁𝗲𝗺𝗮𝘁𝗶𝗰𝗵𝗲, 𝗶𝘀𝗰𝗿𝗶𝘃𝗶𝘁𝗶 𝗮𝗹𝗹𝗮 𝗺𝗶𝗮 𝗻𝗲𝘄𝘀𝗹𝗲𝘁𝘁𝗲𝗿: bit.ly/newsletter-alessiopomaro

    #AI #GenAI #GenerativeAI #IntelligenzaArtificiale #LLM 

  7. A $196 fine-tuned 7B model outperforms OpenAI o3 on document extraction

    arxiv.org/abs/2509.22906

    #HackerNews #A #$196 #fine-tuned #7B #model #outperforms #OpenAI #o3 #on #document #extraction

    fine-tuned-model #document-extraction #OpenAI #AI-research #machine-learning

  8. A $196 fine-tuned 7B model outperforms OpenAI o3 on document extraction

    arxiv.org/abs/2509.22906

    #HackerNews #A #$196 #fine-tuned #7B #model #outperforms #OpenAI #o3 #on #document #extraction

    fine-tuned-model #document-extraction #OpenAI #AI-research #machine-learning

  9. A $196 fine-tuned 7B model outperforms OpenAI o3 on document extraction

    arxiv.org/abs/2509.22906

    #HackerNews #A #$196 #fine-tuned #7B #model #outperforms #OpenAI #o3 #on #document #extraction

    fine-tuned-model #document-extraction #OpenAI #AI-research #machine-learning

  10. A $196 fine-tuned 7B model outperforms OpenAI o3 on document extraction

    arxiv.org/abs/2509.22906

    #HackerNews #A #$196 #fine-tuned #7B #model #outperforms #OpenAI #o3 #on #document #extraction

    fine-tuned-model #document-extraction #OpenAI #AI-research #machine-learning

  11. A $196 fine-tuned 7B model outperforms OpenAI o3 on document extraction

    arxiv.org/abs/2509.22906

    #HackerNews #A #$196 #fine-tuned #7B #model #outperforms #OpenAI #o3 #on #document #extraction

    fine-tuned-model #document-extraction #OpenAI #AI-research #machine-learning

  12. 🧠 Qual è la differenza tra usare la Web Search in agenti dotati di "reasoning" e non? Ad esempio usando GPT-4.1 oppure #o3..

    👉 Alcune riflessioni: linkedin.com/posts/alessiopoma

    ___ 
    ✉️ 𝗦𝗲 𝘃𝘂𝗼𝗶 𝗿𝗶𝗺𝗮𝗻𝗲𝗿𝗲 𝗮𝗴𝗴𝗶𝗼𝗿𝗻𝗮𝘁𝗼/𝗮 𝘀𝘂 𝗾𝘂𝗲𝘀𝘁𝗲 𝘁𝗲𝗺𝗮𝘁𝗶𝗰𝗵𝗲, 𝗶𝘀𝗰𝗿𝗶𝘃𝗶𝘁𝗶 𝗮𝗹𝗹𝗮 𝗺𝗶𝗮 𝗻𝗲𝘄𝘀𝗹𝗲𝘁𝘁𝗲𝗿: bit.ly/newsletter-alessiopomar 

    #AI #GenAI #GenerativeAI #IntelligenzaArtificiale #LLM 

  13. LMArena is losing relevance: general prompts and fine-tuned models skew results.

    Allen AI's SciArena offers a fix: 100+ domain experts evaluate answers grounded in retrieved literature.

    OpenAI's o3 dominates:

    59% wins vs C4 Opus
    80% vs Gemini 2.5 Pro

    Why?

    o3 gives deeper citations, clearer structure, precise terminology, and broader coverage.

    A step forward for expert-led model evaluation.

    allenai.org/blog/sciarena

    #llm #gemini #o3 #claude #chatgpt

  14. 🧠 Un esempio di una chiamata API di #o3 con la Web Search attiva.

    👉 Come funziona: linkedin.com/posts/alessiopoma

    ___ 
    ✉️ 𝗦𝗲 𝘃𝘂𝗼𝗶 𝗿𝗶𝗺𝗮𝗻𝗲𝗿𝗲 𝗮𝗴𝗴𝗶𝗼𝗿𝗻𝗮𝘁𝗼/𝗮 𝘀𝘂 𝗾𝘂𝗲𝘀𝘁𝗲 𝘁𝗲𝗺𝗮𝘁𝗶𝗰𝗵𝗲, 𝗶𝘀𝗰𝗿𝗶𝘃𝗶𝘁𝗶 𝗮𝗹𝗹𝗮 𝗺𝗶𝗮 𝗻𝗲𝘄𝘀𝗹𝗲𝘁𝘁𝗲𝗿: bit.ly/newsletter-alessiopomar 

    #AI #GenAI #GenerativeAI #IntelligenzaArtificiale #LLM 

  15. How to pass an AI coding benchmark: train on the questions

    SWE-Bench Verified by OpenAI tests how well a model can solve real bugs in real Python code from GitHub. These bugs are all public information — so the AI models have almost certainly trained on the actual text of the bug and on the fix for the bug. In “The SWE-Bench Illusion,” researchers at Purdue […]

    pivot-to-ai.com/2025/07/02/how

  16. Freaky Friday auf Ö3! Heute TV-Songs. Da kommen Erinnerungen hoch..😃
    #freakyfriday #Ö3 #Radio

  17. Wieder Freaky Friday auf Ö3!
    Das Thema heute: Insel! 🏝️
    #Radio #FreakyFriday #Ö3

  18. A note about #chatgpt tasks ☝️ 🕐

    Although you (might) no longer see the model with task support in the model selector, you can still create new tasks with chatgpt by just choosing #o3 and typing "create a new task that...."

    And you can still manage your tasks via the browser: chatgpt.com/tasks

    The tasks feature is great. It's sad that @openai puts so little effort into this 😢

    #tip #ai #openai

  19. If you can’t use a billion-dollar #ai system to solve a problem that Herb Simon (one of the actual godfathers of AI) solved with classical (but out of fashion) AI techniques in 1957, the chances that models such as #claude or #O3 are going to reach artificial general intelligence ( #agi ) seem truly remote.

    #tech #technology #science #software #programming #llm #chatbots #research

    theguardian.com/commentisfree/