12daysofopenai.com Open in urlscan Pro
104.21.95.209  Public Scan

Form analysis
1 forms found in the DOM

POST https://script.google.com/macros/s/AKfycbz6VvT3_KJNizZQv4Lp0YovOW304cfrjeBTRcs4P2Lplpe4SlnWqkKlasQhd3yGXEE4oA/exec

<form id="newsletterForm" class="newsletter-form" method="POST" action="https://script.google.com/macros/s/AKfycbz6VvT3_KJNizZQv4Lp0YovOW304cfrjeBTRcs4P2Lplpe4SlnWqkKlasQhd3yGXEE4oA/exec">
  <div class="form-group">
    <input type="email" id="email" name="Email" placeholder="your@email.com" required="">
  </div>
  <div class="form-group">
    <input type="text" id="name" name="Name" placeholder="Your Name" required="">
  </div>
  <button type="submit" class="subscribe-button">Subscribe</button>
  <div id="formMessage" class="form-message"></div>
</form>

Text Content

12 DAYS OF OPENAI

Tracking OpenAI's special announcements and features during their 12 Days event.

> 

> 

Day 1 Day 2 Day 3 Day 4 Day 5 Day 6 Day 7 Day 8 Day 9 Day 10 Day 11 Day 12


DAY 1 RESOURCES

📝 Full Transcript & Summary Complete presentation transcript with timestamps,
detailed summaries, and key highlights


TODAY'S MAJOR UPDATES

🚀 O1 MODEL RELEASE

 * • Chain of Thought reasoning with 50% faster processing
 * • Error rate reduced by 34%
 * • Advanced multimodal capabilities surpassing GPT-4o
 * • 93% accuracy on bias evaluations
 * • Hallucination rate reduced to 44% (vs GPT-4o's 61%)
 * • Superior performance across 14 languages

💎 CHATGPT PRO ($200/MONTH)

 * • Unlimited access to O1 and O1-Pro mode
 * • Enhanced performance on complex tasks
 * • 20% improvement in knowledge tests
 * • Unlimited advanced voice mode
 * • Optimized for development and writing

🛡️ SAFETY & PERFORMANCE

 * • 100% standard safety test pass rate
 * • 92% on challenging safety tests (vs GPT-4o's 71%)
 * • 84% pass on StrongReject jailbreak tests
 * • Clear hierarchy: System > Developer > User
 * • Risk Assessment: Medium for CBRN/Persuasion, Low for Cybersecurity/Autonomy

📄 O1 System Card 💎 ChatGPT Pro Announcement


DAY 1 PRESENTATION




DAY 2 RESOURCES

📝 Full Transcript & Summary Complete presentation transcript with analysis of
Reinforcement Fine-tuning demonstration


TODAY'S MAJOR UPDATES

🔄 REINFORCEMENT FINE-TUNING (RFT)

 * • Advanced from "high school to PhD level" reasoning
 * • Achieves significant results with just 12 examples
 * • Same technique used in GPT-4 and O1 series training
 * • Public launch scheduled for early 2025
 * • Implemented in Thompson Reuters' Co-counsel AI

⚙️ TECHNICAL IMPLEMENTATION

 * • JSONL format training data structure
 * • Separate validation dataset prevents memorization
 * • Advanced grading system (0-1 score range)
 * • Distributed training infrastructure support
 * • Automated hyperparameter optimization

🧬 BERKELEY LAB CASE STUDY

 * • 300 million people affected by rare diseases globally
 * • 1,100 training examples from medical literature
 * • 31% Top-1 accuracy (Fine-tuned O1 Mini)
 * • Surpassed base O1 performance (25%)
 * • Novel open-ended symptom querying capability

🎥 Demo Video 🔬 RFT Research Program Application 💬 HN Discussion 📚 Justin
Reese Research


DAY 2 PRESENTATION


🔜


COMING ON DECEMBER 7TH

Stay tuned for December 7th's announcement!

🔜


COMING ON DECEMBER 8TH

Stay tuned for December 8th's announcement!

🔜


COMING ON DECEMBER 9TH

Stay tuned for December 9th's announcement!

🔜


COMING ON DECEMBER 10TH

Stay tuned for December 10th's announcement!

🔜


COMING ON DECEMBER 11TH

Stay tuned for upcoming announcement!

🔜


COMING ON DECEMBER 12TH

Stay tuned for upcoming announcement!

🔜


COMING ON DECEMBER 13TH

Stay tuned for upcoming announcement!

🔜


COMING ON DECEMBER 14TH

Stay tuned for upcoming announcement!

🔜


COMING ON DECEMBER 15TH

Stay tuned for upcoming announcement!

🔜


COMING ON DECEMBER 16TH

Stay tuned for upcoming announcement!


STAY UPDATED

Subscribe to receive the latest updates and announcements from OpenAI's 12 Days
event directly in your inbox.


Subscribe


Not affiliated with OpenAI. This is a community project tracking the 12 Days
event.