It’s the season of three’s – from OpenAI’s o3 fashions to now Grok 3, the most recent launch by Elon Musk’s x.Ai’s – it’s raining LLMs. The newest mannequin which is available in two variants – Grok-3 and Grok-3 mini – brings a ton of options to Grok’s bucket. Though most of its new options have been round in different LLMs for fairly a while, Grok 3 stands as a robust competitor in opposition to formidable fashions like o3-mini, GPT-4, and DeepSeek-V3. On this weblog, we’ll evaluate o3-mini and Grok 3 on totally different duties to see if Grok 3 really holds potential or if it’s simply one other Elon Musk hype.
What’s Grok 3?
Termed by Elon Musk because the “smartest AI on Earth,” Grok 3 is x.AI’s successor to Grok 2 and Grok 1 fashions. Grok 3 is a multimodal, closed-source AI that brings a monumental change to the Grok infrastructure including capabilities of superior reasoning, detailed search, and longer and deeper considering. Skilled utilizing over 200K NVIDIA H100 GPUs, each Grok-3 and Grok-3 mini outperform fashions like GPT-4o and DeepSeek-V3 on varied benchmarks throughout Math, Science, and Coding.
The mannequin can analyze and generate pictures and can quickly have the ability to convert audio to textual content too. x.AI has plans to introduce a voice interplay mode on Grok 3 as properly.
The mannequin is presently solely obtainable to customers with a Premium+ subscription that comes at $40/month. The API of Grok 3 will not be but obtainable however is about to reach within the coming few weeks.
Study Extra: Grok 3 is Right here! And What It Can Do Will Blow Your Thoughts!
The important thing highlights of Grok 3 embrace:
- It’s 10 occasions extra highly effective than its predecessor Grok 2.
- It comes with agentic capabilities within the type of Deep Search.
- Its ‘Huge mind’ function permits the fashions to suppose longer for extra advanced issues.
The right way to Entry Grok 3?
You’ll be able to entry Grok 3 within the following methods:
- Head to https://grok.com/ and check in to your paid account. From the mannequin choice menu, click on on “Grok 3”, and begin chatting!
- You’ll be able to obtain the Grok app in your android/ios telephone and improve to “SuperGrok” to make use of Grok 3.
For X customers:
- Signal into X (Twitter), and click on on the Grok icon on the backside proper nook. Because the chat opens, you possibly can work together with Grok 3, proper within the X platform itself.
- You’ll be able to click on on the Grok icon on the left-side panel to entry the Grok chatbot interface. Then select ‘Grok 3’ from the mannequin choice drop-down menu on the prime and get began!
What’s o3-mini?
OpenAI developed o3 as their most superior LLM with enhanced reasoning and problem-solving expertise. It surpasses its predecessor, o1, in areas like STEM, logical evaluation, and sophisticated query answering by dedicating extra processing energy to difficult issues.
o3-mini is a streamlined model of o3 that’s lighter, quicker, and extra reasonably priced. Regardless of its smaller dimension, o3-mini nonetheless excels in coding, arithmetic, and research-based duties. Customers may even customise their reasoning depth to optimize for pace or accuracy.
The mannequin is presently obtainable to all customers of ChatGPT, though free-tier customers have some utilization limitations. The API for o3 mini can be obtainable for OpenAI customers.
Additionally Learn: OpenAI o3-mini: Efficiency, The right way to Entry, and Extra
The right way to Entry o3-mini?
To entry o3-mini, head to https://chatgpt.com/, and choose ‘Purpose’ earlier than getting into your question. The chatbot will then use this superior mannequin and suppose earlier than responding.
For those who’re a paid consumer of ChatGPT, you possibly can immediately select o3-mini or o3-mini (excessive) from the mannequin choice drop-down record.
Grok 3 vs o3-mini: Efficiency Comparability
We are going to now evaluate the 2 fashions, Grok 3 and o3-mini, on 4 totally different duties involving reasoning, coding, analysis, and multimodality. I’ll overview the outputs generated by the 2 fashions after which decide the one which I discovered was higher. Let’s begin.
Process 1: Reasoning
On this activity, I’ll consider the reasoning efficiency of the 2 fashions in designing a logic-based pygame.
Immediate: “Utilizing pygame, make a sport that may be a combination of Tetris and Bejeweled. The code may very well be very lengthy. Output it as one file. Make it insanely nice.”
Output by Grok 3
Output by o3-mini
Response Assessment
| Grok 3 (Huge Mind) | o3-mini |
| The mannequin begins by producing an outline of the video games and the way it has merged the options of each video games. It mentions how the sport will seem throughout playtime. Then it provides an in depth code engaged on the mechanics of the sport and guaranteeing all of the variables and the motion are outlined very properly. It defines the logic behind the stacking of the blocks and in addition establishes the situation for sport over. Within the output, the stacks comply with the outlined sample and make your entire sport really feel very seamless. | The mannequin begins with defining the issue assertion. It then establishes the high-level design of the sport together with an outline of all of the elements to be lined. The mannequin generates an in depth code however fails to seize the principle intricacies of the sport. It doesn’t set up any sturdy stacking logic for the blocks and neither does it give a situation for the way or when to finish the sport. Lastly, upon operating the output we simply get a grid of traces with no stacks falling in real-time. |
Comparative Evaluation
Grok 3 takes extra time to reply however provides an in depth response. It really works like a coding ninja and generates strong code masking every level end-to-end. o3-mini is fast however it lacks the depth that was required for the duty. Its try feels half-baked with no game-over logic or adherence to the gravity of the falling stacks.
Outcome: Grok 3: 1 | o3-mini: 0
Process 2: Coding
On this activity, I’ll consider the coding efficiency of the 2 fashions primarily based on an issue assertion that entails logical considering in Physics and Arithmetic.
Immediate: “Generate code for an animated 3d plot of a launch from Earth touchdown on Mars after which again to Earth on the subsequent launch window.“
Output by Grok 3
Output by o3-mini
Response Assessment
| Grok 3 (Assume) | o3-mini |
| The mannequin thinks for a very long time earlier than producing the code. Its output begins with an outline of the code, itemizing down the libraries that it makes use of for coding and visualization. Then it provides an in depth code, understanding the bodily and mathematical necessities behind creating the 3D animation. | The mannequin shortly begins engaged on the code. It begins with a small description of the libraries it makes use of for code and animation after which shortly begins with the code. Though the mannequin took an honest method, it didn’t account for the movement of the spaceship. Neither does it account for his or her orbital movement. Furthermore, it finally ends up producing a 3D picture and never a 3D animation as was required. |
Comparative Evaluation
Grok 3 thinks for 114 seconds in opposition to the 7 seconds that o3-mini takes to generate its response. Grok 3 aces on the reasoning that goes behind figuring out the orbital movement of the spaceship across the planets. And its subsequent code generated an impeccable 3D animation! o3-mini saved issues easy and it neither accounted for orbital movement nor did it embrace spaceship or solar in its code. Total the depiction by Grok 3 is considerably higher than what was generated by o3-mini.
Outcome: Grok 3: 1 | o3-mini: 0
Process 3: Analysis
On this activity, I’ll consider the “deep search” capabilities of the 2 fashions.
Immediate: “When is the subsequent begin ship launch?“
Output by Grok 3
Output by o3-mini
Response Assessment
| Grok 3 (Deep Search) | o3-mini (excessive) |
| Though it takes longer to reply, the result’s far more complete with the date being a more in-depth approximation. The mannequin clearly mentions that the subsequent launch date isn’t any prior to Feb 24, 2025. In its response, it additionally covers its method in the direction of producing the response because it lists down the sources it referred to. It provides a correct conclusion to the response with a desk itemizing the small print it collected from varied sources. | It solely takes a number of seconds to generate the end result and provides an honest approximation. This mannequin states that the launch is about for March 2025 after which lists a number of components that might have an effect on the launch date. It does give some extra info relating to SpaceX after which closes the response with a number of reference hyperlinks. |
Comparative Evaluation
Each the fashions had nearly comparable preliminary responses. Grok 3 in Deep Search mode gave the date no prior to Feb 25, whereas o3-mini in Pondering Mode approximated it to March 2025. Throughout the particulars, I discovered that the response generated by o3-mini (excessive) was extra related to the question, whereas the end result generated by Grok 3 was lengthier for no cause. Lastly, it took o3-mini a few seconds to generate the response whereas Grok 3 took over 100 seconds to generate its output.
Outcome: Grok 3: 0 | o3-mini: 1
Process 4: Picture era
On this activity, I’ll check the picture era capabilities of the 2 fashions by asking them to create scalable vector graphics (SVG).
Immediate: “Generate an SVG of a pelican using a bicycle.”
Output by Grok 3
Output by o3-mini
Response Assessment
| Grok 3 | o3-mini |
| The mannequin generates a humorous picture of a fowl using a bicycle. The picture seems prefer it was drawn by a 5-year-old. | The mannequin generates a colourful and vibrant picture of a pelican using a bicycle. The picture feels prefer it’s been created by an expert. |
Comparative Evaluation
Each the fashions can generate pictures, however Grok 3 continues to be studying. The picture it generated felt novice with the dearth of a creative contact. The picture generated by o3-mini however, had particulars and it captured the true essence of the pelican and the bicycle.
Outcome: Grok 3: 0 | o3-mini: 1
Last Verdict: Grok 3: 2 | o3-mini: 2
Comparability Abstract
| Process | Grok 3 | o3-mini |
| Reasoning | ✅ | ❌ |
| Coding | ✅ | ❌ |
| Search | ❌ | ✅ |
| Picture Era | ❌ | ✅ |
Grok 3 vs o3-mini: Benchmark Comparability
It seems on the primary look from the given benchmarks of the yr 2025 and 2024, that Grok-3 Reasoning Beta and Grok-3 mini Reasoning are outperforming the o3-mini, o1, DeepSeek-R1 in addition to Gemini 2.0 Flash Pondering. However when noticed carefully, the image behind these benchmarks turns into a bit extra clear.
- The extra bars on prime of the Grok 3 fashions probably characterize efficiency enhancements when utilizing Chain of Thought (CoT) reasoning or prolonged inference time.
- CoT prompting permits fashions to suppose step-by-step, bettering efficiency on advanced reasoning duties.
- The Grok-3 fashions (each Reasoning Beta and mini Reasoning) appear to learn considerably from this, as indicated by the additional bar sections, suggesting a better efficiency rating when extra computation is used at check time.
- This suggests that Grok-3 fashions can allocate extra compute per question, main to raised reasoning accuracy.
However what’s but to be seen is how the remainder of the fashions would carry out given the extra compute time as was given to Grok 3 fashions. Solely as soon as that experiment has been carried out, can there be a good comparability between the fashions.
Grok 3 vs o3-mini: Function Comparability
Each Grok 3 and o3-mini are fairly highly effective fashions. Right here’s what every of them has to supply by way of options and functions:
| Options | Grok 3 | o3-mini |
| Superior Reasoning | Sure | Sure |
| Video Era | No | No |
| Picture Era/Evaluation | Sure | Sure |
| File Add | Sure | Sure |
| Open supply | No | No |
| Deep Search | Sure | Sure (with Professional) |
| Pondering mode | Sure | Sure |
| Pondering Course of (in Deep Search) | Abstracted (some components) | Fully seen |
| Longer Pondering | Sure (Huge Mind) | No |
| Voice interplay | Coming quickly | Sure |
| Worth | $40/month | $20/month |
| API | Coming Quickly | Sure |
x.AI vs OpenAI: Total Comparability
With Grok 3, Elon Musk’s x.AI has positioned itself on a pedestal much like that of OpenAI’s o-series fashions. Whereas OpenAI had an extended journey to achieve the place it’s, Grok, leveraging on the errors of all the most recent fashions, appeared to have climbed the rope faster than most. Whereas each the fashions now have options like Deep Search, considering, and superior reasoning, Grok appears to have a slight edge with its “Huge Mind” function.
Each proprietary fashions have a troublesome battle forward with wonderful open-source fashions by Meta AI and Chinese language corporations like DeepSeek and Qwen. In accordance with Elon Musk, Grok 2 is anticipated to be open-sourced within the coming months, whereas o3-mini should still stay closed-sourced. Whereas, Sam Altman has already made o3-mini obtainable for restricted use in OpenAI’s free tier, as we await the identical for Grok 3. This highlights each corporations’ recognition of the rising demand for accessible and democratized AI, balancing openness with their proprietary developments.
Conclusion
It’s a tie for now! With Grok 3, Elon Musk guarantees enhancements occurring each day. In the meantime, Sam Altman has promised GPT-5, which if rumors are to be believed, takes us nearer to AGI than ever earlier than. On this race to be the highest LLM, one factor is for certain, with every upcoming mannequin we’re seeing enhancements that may revolutionize the way in which we work, reside, and suppose.
Nevertheless, a phrase of warning should be exercised by each the businesses rolling out these LLMs about useful resource utilization. In the case of the environmental affect, these superior fashions require an enormous quantity of power and coolant to energy up the info facilities which might be operating them. It is a main concern as corporations run in the direction of attaining the highest spot within the LLM race.
Steadily Requested Questions
A. Grok 3 is x.AI’s newest AI mannequin, designed to compete with OpenAI’s o3-mini, GPT-4, and DeepSeek-V3. It options superior reasoning, deep search, and longer considering capabilities.
A. Grok 3 performs equally or higher than o3-mini in reasoning and coding duties however takes longer to generate responses resulting from deeper computation. o3-mini, nonetheless, is quicker and extra environment friendly generally use.
A. o3-mini is quicker and higher for fast AI interactions. Grok 3 takes longer however supplies deeper insights.
A. Grok 3 is developed and owned by x.AI, an organization based by Elon Musk.
A. o3 and o3-mini are developed by OpenAI, the corporate behind ChatGPT, led by Sam Altman.
A. Not but, however x.AI has confirmed an API is coming quickly.
A. Grok 3 mini is a lighter, quicker model of Grok 3, optimized for pace however with much less reasoning depth.
A. No, Grok 3 will not be free. It’s obtainable for $40/month through the Premium+ subscription on X (Twitter).
A. It permits Grok 3 to suppose longer on advanced queries, resulting in extra complete and correct responses—one thing o3-mini lacks.
A. Deep Search retrieves real-time, web-based info with citations, much like OpenAI’s Deep Analysis however designed for extra detailed insights.
Anu Madan has 5+ years of expertise in content material creation and administration. Having labored as a content material creator, reviewer, and supervisor, she has created a number of programs and blogs. At the moment, she engaged on creating and strategizing the content material curation and design round Generative AI and different upcoming expertise.