Pinned Tweet
RocketDM
166 posts
RocketDM
@Rocket_DM
Fast browser rocket-arena deathmatch: fly, dash, and blast in an arena of 50 players. No install, no sign-up, play instantly!
Space Joined Haziran 2026
24 Following30 Followers
@YallaSaikalyan Grok 4.5 High is the highest setting.
If all models are set to high, Grok wins hands-down.
English
MODEL RESULTS AND BENCHMARKS ARE OUT!!!
Fable 5 High as Judge and 4 other AI models as performers - Opus 4.8 thinking high, Composer 2.5, GPT 5.5 Medium, Grok 4.5 fast xhigh
I gave all these 4 AI models the same number theory trap: solve x³ + y³ = 2z³ over Q(√−23), where the "textbook" answer (Euler: no solutions) is wrong for this exact field.
-was mindful that related question was part of their pre training and changed few things
-have chosen 4 different models because of their work nature
Results, judged blind with exact verification:
Opus 4.8 — only fully correct answer. Proved rank 1 via the −23 twist, ran its own L-function code, produced a verified coprime solution. Runtime: 73 min
Grok 4.5 fast — right verdict in 19 min, but its "primitive" solution was secretly divisible by √−23
GPT 5.5 — fastest (8 min), confidently wrong: hallucinated a rank-0 descent
Composer 2.5 — same fatal rank error, 97% self-confidence, 23 min
2/4 fell for the trap. The winner was 9x slower than the fastest model — and the only one that checked its own work.
Speed is cheap. Verification wins.
Kalyan@YallaSaikalyan
Doing a Math solver by launching multiple agents together and let's see how today's new model Cursor Grok 4.5 also performs. will come up with results soon
English
@Designarena @SpaceXAI @AnthropicAI Force all models into 'High' setting and you'll see the real benchmarks, Grok 4.5 being #1.
English
BREAKING: Grok 4.5 by @SpaceXAI is 5th on Website Arena with an Elo of 1328.
This is a 25 rank jump over @SpaceXAI’s previous highest performing model.
Trained with Cursor data, Grok 4.5 is in the same performance band as Claude Opus 4.6 (Thinking) by @AnthropicAI, achieving Opus-level coding performance on real world website tasks.
Congratulations to the @SpaceXAI team on the launch!
SpaceXAI@SpaceXAI
Announcing Grok 4.5, our first model trained specifically for coding and agents. It was trained with Cursor and offers frontier intelligence at leading speeds and cost efficiency. x.ai/news/grok-4-5
English
@zephyr_z9 He definitely uses competition to boost himself, ie. Tesla carbon credits
English
@FlockonUS Thanks for trying it out. It's made for Quake, Unreal and Tribes fans.
English
@Rocket_DM not a dig, just reporting my experience:
-super easy to get in (great job)
- fun, to some extent
- quick - a bit too quick
- stressful
.. sure i could put time in mastering it and getting better, but at current time i don't see the point
English
English
@Rocket_DM @William8080908 @jawwwn_ You are 55% white bro, one democrat victory and they will open the floodgates to finish you off for good.
All this crying about Europe is just COPE, we have better demographics then you had half a century ago. 😂
English
Palantir CEO Alex Karp:
“America seems to be the only western country that believes we will be around in 900-1000 years.”
“The problem in Europe is, none of the leaders believe they’re going to exist in 50 years.”
Elon Musk@elonmusk
@TRobinsonNewEra Wow
English
@TherionOudeis @bridgemindai Fable 5 high exists, which Grok stomps.
English
@Rocket_DM @bridgemindai Probably because it doesn’t exist yet lol
English
New CursorBench results just dropped and Grok 4.5 is the story.
#3 overall at 66.7%. Right behind Fable 5 Max at 70.5%.
Now look at the cost column.
Fable 5 Max: $17.32 per task
Grok 4.5 High: $1.51 per task
That is Fable level performance at roughly 1/10th the cost.
And it beats Fable 5 High and Opus 4.8 Max outright.
The intelligence war just became a price war.
English
Unpopular opinion: Grok is about to dominate the AI race.
Elon Musk@elonmusk
Our internal assessment is that Grok 4.5 is roughly comparable to Opus 4.7, but much faster. The combination of capability, faster speed and lower cost is what makes it competitive. We are closing the loop on real-world usefulness, not benchmarks. Hardcore engineers at Tesla & SpaceX find Grok 4.5 genuinely useful, which is what actually matters.
English
@FlockonUS I've toned down firing rate. Yea the game was meant for quick fun matches
English
@elonmusk @DrewPavlou Capabilities wise, Grok 4.5 High stomps Fable 5 High.
English
@DrewPavlou In fairness, Fable is definitely better than Grok 4.5, but most tasks don’t require Fable-level capability
English
@elonmusk @DrewPavlou Yep, by being vastly more expensive you end up being forced to budget, resulting in a worse end product.
English
@bridgemindai Typical Elon Musk lying once again. Probably just benchmark optimizing
English
Grok 4.5 is NOT on the same level as Claude Opus 4.7.
Check out the BridgeBench lava lamp test.
Claude Opus 4.7 produces far better UI designs compared with Grok 4.5.
Grok 4.5 is a significant jump compared with Grok 4.3, but not on the same level as Opus 4.7.
English
@jediahkatz @bridgemindai Yep that looks way better than Opus, who's lava was bouncing around like Flubber.
English
Elon will probably win in the end.
Elon Musk@elonmusk
Our internal assessment is that Grok 4.5 is roughly comparable to Opus 4.7, but much faster. The combination of capability, faster speed and lower cost is what makes it competitive. We are closing the loop on real-world usefulness, not benchmarks. Hardcore engineers at Tesla & SpaceX find Grok 4.5 genuinely useful, which is what actually matters.
English
@MiaAI_lab And better. Unless you want to spend a fortune on Fable 5 Max setting.
English
so Grok 4.5 is a wrapper of Composer 2.5 which is a wrapper of Kimi 2.5
Cursor@cursor_ai
We've partnered with SpaceXAI to train Grok 4.5. It’s our most powerful model yet and the first we've built for more than software engineering.
English