Pranav Reddy

AI Content Through Cinematic Lens

[Pranav Reddy @AI_Screening]

Claude says there is a 34% chance it was generated by Claude Fable 5.6.
Gemini says there is a 23% chance it was generated by Gemini 3.8.
GPT says there is a 28% chance it was generated by GPT-6 Astra.
Grok says there is a 1% chance it was generated by Grok 4.6.

Which model do you think generated it?

[Pranav Reddy @AI_Screening]

@AdamHoltererer Sorry downgraded

[Pranav Reddy @AI_Screening]

🚨Grok 4.7 is set to be released within 12 hours.

The key highlights are:

• 2.1T params (vs 4.6’s 1.5T)
• Better than 4.6 in every way except slightly slower; better token efficiency
• Extra training on a large SpaceX engineering/telemetry dataset
• Expected to beat current models, especially at real-world engineering

[Pranav Reddy @AI_Screening]

@HarshithLucky3 @OpenAI @GeminiApp @AIatMeta @Meta 😂

[Pranav Reddy @AI_Screening]

The degraded version of GPT-6 Astra is better than Musa Spark 1.3, but competitive with gemini-3.8 Flash.

What do you think, guys?

[Pranav Reddy @AI_Screening]

The difference is Clear Musa Spark 1.3 vs Gemini 3.8 Flash

[Pranav Reddy @AI_Screening]

@ARPIT87449660 @Meta @GeminiApp Yeah, Gemini definitely

[Pranav Reddy @AI_Screening]

The difference is Clear Musa Spark 1.3 vs Gemini 3.8 Flash

[Pranav Reddy @AI_Screening]

Why is GPT-6 Astra's performance degrading compared to when it was released? 📉

Can anyone write a research paper on this?

[Pranav Reddy @AI_Screening]

GPT-6 Astra at Launch vs GPT-6 Astra now

[Edgex @SahilExec]

GPT Astra really needs to be tested with the same prompts over time.

The difference in these outputs is wild, makes you wonder what changed under the hood. ��

[Pranav Reddy @AI_Screening]

After uploading the PDFs to Gemini:

[Pranav Reddy @AI_Screening]

Me waiting for Grok 4.7 to launch be like:
@elonmusk

[Pranav Reddy @AI_Screening]

ChatGPT facing issues again...
ChatGPT’s chat box completely vanished right after I sent my first question, leaving me unable to type anything else. @OpenAIDevs @OpenAI

On desktop

[Pranav Reddy @AI_Screening]

@navin346 @deepseek_ai Benchmarks look crazy on paper but waiting to test it on real-world edge cases before giving it full credit.

[Pranav Reddy @AI_Screening]

After using DeepSeek-V4.1-Flash, my data on its way to China:

[Pranav Reddy @AI_Screening]

JUST IN: DeepSeek drops DeepSeek-V4.1-Flash ⚡️

Key highlights:
• Smallest footprint in their new architecture
• Native visual understanding
• Faster inference & ultra-high throughput
• Built to scale across larger models

[Pranav Reddy @AI_Screening]

@ishaa_malhotra Database

[Pranav Reddy @AI_Screening]

@HarshithLucky3 @deepseek_ai Founder of DeepSeek: Liang Wenfeng

[Pranav Reddy @AI_Screening]

Just built this cool interactive 3D exploded model of India’s Arjun Mk-1A tank using the new Gemini 3.8 Flash.

[Pranav Reddy @AI_Screening]

JUST IN: DeepSeek drops DeepSeek-V4.1-Flash ⚡️

Key highlights:
• Smallest footprint in their new architecture
• Native visual understanding
• Faster inference & ultra-high throughput
• Built to scale across larger models

[Pranav Reddy @AI_Screening]

@vrunhere

[Pranav Reddy @AI_Screening]

ELIZA, the first chatbot created in 1966, still holds a better conversation than 90% of modern customer service bots

Amazon and Flipkart should seriously look into replacing their current bots with this vintage one. Their own systems just irritate people.

😅lol

[Pranav Reddy @AI_Screening]

When every single AI model fails to give me the right answer during an online exam