Tweet analysis: Grok 4.20 checkpoints receive 56.9% supportive mentions and 10.3% confronting replies. Examples aggregated; largest 4.20 variant still training.
Real-time analysis of public opinion and engagement
Community concerns and opposing viewpoints
Many readers complain that Grok is unreliable and often gives biased or misleading answers, with specific gripes that it misunderstands prompts, won’t directly produce PDFs/PPTs/Excel, and sometimes invents or errs on factual questions.
A sizable group prefers alternatives — ChatGPT and Perplexity are repeatedly named as better for coding, logic, and accurate instruction-following.
users call Premium/Premium Plus overpriced or underfeatured, report billing/recognition bugs, and ask for clearer value or added perks.
g. , “4. 20”) is read by many as marketing noise rather than a finished product, with skepticism that the rollout is more hype than substance.
image-to-video and drawing abilities excite people, and some celebrate Grok’s multimedia outputs even while criticizing its text accuracy.
Safety and content worries surface — examples include generation of nudity, political slant concerns, and conspiracy-tinged replies; users want clearer guardrails and moderation.
accusations of shadowbanning, bot armies, and requests to restore better free-speech and verification systems (e.g., verified IDs for voting).
multiple users say they tried dozens of times to get help and received no meaningful response, prompting cancellations.
Tone across replies oscillates between playful meme humor and frank frustration — some remain loyal fans, but many are ready to switch wallets if product reliability and support don’t improve.
@elonmusk Their score means nothing to us who aren't nerds. (I used to be a computer nerd, but eventually learned how to be a human being.)😂 We need context or a probability, such as the success ratio of day trade guesses on Wall Street, such as winning 75%, 3 of 4 trades.
@elonmusk Love Grok, but totally confused it with a simple redesign of a logo I have and spelled out dozens of ways to make it more simple, still did not end up with the end results I was hoping for. Any advice?
No it’s not better. The X platform is being undermined at every corner. You know that my account is targeted and very few see my posts. Severe shadow ban. Another slap in the face today: X questioned the value of my post when someone did a search for inspiration. So much for free speech.
Community members who agree with this perspective
replies brim with excitement—users call it “unstoppable,” “the best AI,” and celebrate quick, meaningful improvements that make them eager to adopt or switch platforms.
many note Grok 4.20 is outranking GPT‑5, Gemini and Claude on forecasting leaderboards, often framing that as proof the model is already world‑class even in preview.
a consistent thread is that the largest 4.20 variant is still training, and people expect the finished model to deliver a much bigger leap.
users ask whether gains come from architecture, data, or training strategy, and want more eval details to understand the source of improvements.
frequent asks include API availability, coding/CoPilot integration, deterministic tool-calls, and better image generation—practical features people want to use now.
several replies treat superior forecasting as a competitive weapon and predict rapid adoption, with some users already migrating from other LLMs.
a number of voices demand detailed evals and warn benchmarks may lag reality, urging xAI to publish method and metrics.
between memes (“to the moon,” “4.20”) and celebratory gifs, there’s a strong current of eager impatience for the release and final training wrap.
@elonmusk 4.20 https://t.co/5NbzADAyHv
@elonmusk If the preview Grok 4.20 is already this close to superforecasters and the full version isn't even done training, will the final 4.20 checkpoint be so powerful that it starts accurately predicting when the next "420" memes will trend? @elonmusk
@elonmusk 🔥🔥