We ’re now into the third year of the AI boom , and manufacture loss leader are showing no signs of slowing down , pushing out new and ( presumably ) more capable model on a steady fundament . ChatGPT , of course , remains the undisputed leader .
But with more than a half - dozen models usable from OpenAI alone , figuring out which one to use for your specific project can be a intimidating task .
o1
As a reasoning fashion model , o1 is touted as using anthropomorphic reasoning to check its own answers before output more exact reception to complicated scientific discipline , math , and coding queries . aim on a “ totally Modern optimization algorithm and a unexampled preparation dataset specifically tailored for it , ” per the company , o1 drastically outperform the GPT-4ofamily of models across industriousness benchmark tests .
The full version of OpenAI o1 made its debut in December 2024 during the company ’s 12 Days of OpenAI livestream effect . Its inner system of checks reportedly help the new framework hallucinate less than its predecessor . According to OpenAI ’s tests , the Modern edition of o1 reduce “ major errors ” on “ difficult real - man questions ” by 34 % compared to its prevue version . If you ’ve catch ream of extremely complex data point to pour over or have complicated mathematics and put one across problems to solve , o1 is the OpenAI modelling you ’ll want to use .
Unfortunately , it is not currently available to free - tier users . you could access o1 with a Plus or Teams subscription , though you ’ll only get 50 messages with it per week at that degree . You ’ll need to shell out $ 200 monthly for aChatGPT Pro subscriptionif you desire to use it without limit . So , our recommendation would be to get down out on 01 , and if you run out of your allotted messages for the 24-hour interval , move on to one of the models below .
o1-mini
OpenAI ’s o1 - mini is a lightweight version ofthe society ’s larger reasoning good example . It is geared more toward rapidly resolve challenging mathematics and razz problem rather than creative prose generation . It ’s built for speed anddesigned to cater fast , efficient solution for specific technical tasks , making it ideal for job mechanisation .
03
o1 enjoyed its position at the vertex of OpenAI ’s model mess for less than a fortnight . On the twelfth and final livestream of the company’s12 Days of OpenAIevent , CEO Sam Altmanrevealed the o1 ’s successor manikin , knight o3 to avoid a right of first publication conflict with a UK telecom .
o3 , our late reasoning fashion model , is a find , with a footmark function betterment on our hard benchmarks . we are commence rubber examination & amp ; red teaming now.https://t.co/4XlK1iHxFK
& mdash ; Greg Brockman ( @gdb)December 20 , 2024
Billed as a “ find , with a step function advance on our hardest benchmarks , ” by OpenAI President and Co - Founder , Greg Brockman , the o3 model deal the same reasoning computer architecture as the o1 house . It just does it better . The o3 model enable exploiter to choose between humble , medium , and gamy compute imagination allocations , with high compute returning the most complete result , albeit at the monetary value of inference time and financial monetary value . ARC - AGI conscientious objector - creator Francois Chollethas notedthat high compute reportedly will be G of dollars per task .
The o3 ’s carrying out is nothing to sneeze at . According to OpenAI , o3 beats out the o1 ’s performance by nearly 23 percent head on the SWE - Bench Verified coding test , more than 60 points gamey on the Codeforce benchmark , and missed just one doubtfulness on theAIME 2024 math test . It also reportedly outperform human experts on the GPQA Diamond test and correctly answered a fourth part of the enquiry on the near impossible EpochAI Frontier Math bench mark where ceremonious LLMs scarce get 2 % of them right .
The o3 poser is presently available in early access to prophylactic and security researchers . It ’s expected to make its public debut in early 2025 .
o3-mini
OpenAI’so3 - mini modelserves the same purpose as o1 - mini or GPT-4o - miniskirt : to serve as a lightweight and more efficient version of the prominent full simulation . declare in tardy January 2025 , o3 - mini is expected to be released as both an API and as an available ChatGPT model “ in a couple weeks,”Sam Altman twit in a post on X.
“ While o1 remains our broader oecumenical - knowledge abstract thought good example , o3 - mini provide a specialized alternative for technical domains necessitate precision and speed , ” the company write in itsannouncement post . “ The release of o3 - miniskirt marks another footstep in OpenAI ’s mission to campaign the boundaries of cost - efficient intelligence operation . ”
GPT-4o
GPT-4ois to GPT-4 in terms of performance whatGPT-4 was to GPT-3.5 . put in in May 2024 , GPT-4o ( the “ o ” stand for “ omni ” ) isOpenAI’s“latest , fastest , highest intelligence model . ” It offers human - level reply prison term ( critical for features likeAdvanced Voice Mode ) , improved performance with oral communication beyond English , and a far well understanding of vision and audio message .
Not only is this poser quicker , more effective , and more monetary value - effective than its predecessors , but 4o has also achieved state - of - the - art result in multilingual and vision benchmark tests . It also offers a 128,000 - role context window and can generate school text , images , audio , and electronic computer computer code . It latterly received a minor rise to itscreative writing skillsthanks to a pre - Thanksgiving update .
you’re able to get at GPT-4o through thefree tier of ChatGPT , but you will turn tail into restrictive usage caps . Once you reach your limit with 4o queries , the platform will send packing your accession down to 4o - miniskirt . support to the $ 20 - per - calendar month Plus level will significantly increase that usage cap .
GPT-4o mini
OpenAI ’s “ lightest - system of weights intelligence model ” isGPT-4o mini . Designed more for make out computationally dewy-eyed tasks performed promptly and repeatedly as oppose to pondering over challenging coding or analytical trouble , 4o mini offers well-nigh all of the same feature of speech as the large 4o model , preserve for its restricted memory access to some of the more advanced analytic prick .
GPT-4o mini is uncommitted to Plus and Pro indorser , and you may select it from the drop - down model listing in the upper left corner of the ChatGPT home screen door . you’re able to also access it as a free - tier exploiter by blowing through your GPT-40 usage allowance . The political platform will shunt you over to the 4o mini exemplar until you come off cool down and are allowed to get at GPT-4o again .
GPT-4
GPT-4was the first great step up in ChatGPT ’s capabilities since its debut . Launched in April 2023 alongside the newChatGPT Plussubscription tier , GPT-4 power both ChatGPT and Microsoft ’s free co-pilot platform . OpenAItouted GPT-4 as“more reliable , creative , and capable to handle much more nuanced instructions than GPT-3.5 . ” It was also the company ’s first multimodal model , allowing it to generate and analyze epitome and audio in addition to text .
GPT-4 still serve as the base model available for barren - level ChatGPT user . For most casual users , GPT-4 and its “ omni ” strain are enough capable of performing the inference tasks that you call for .
Early versions
GPT-1 , the original large terminology mannikin from OpenAI , made its introduction in June 2018 and became the archetype of reproductive pre - trained transformer technology . Even in its early grade , GPT-1 manage to outgo the DoS - of - the - art models of the time by nearly 6 % at natural language inference tasks .
Its 1.5 - billion - parametric quantity heir , GPT-2 , get in November 2019 . It expanded upon GPT-1 ’s basic textual matter generation capableness to let in answer questions , summarizing documents , and translate text between languages . GPT-3 arrived in May 2020 but was speedily snatched up by Microsoft that September with a licensing strategy that gives Microsoft exclusive use of that model .
The GPT-3.5 model is a subset of Microsoft ’s license GPT-3 that offer better public presentation and a later noesis shortcut date than earlier iteration . When OpenAI launched the ChatGPT platform in November 2022 , the chatbot run atop the GPT-3.5 example . The company comprise vane browsing capability the following April , but by May 2023 , GPT-3.5 had been depreciated in favour of the far more capable GPT-4 family of models . As such , it ’s no longer uncommitted for public use and can only be access through OpenAI ’s developer API .
With so many AI chatbots useable , it is unmanageable to take one to use . We rounded up thebest AI chatbotsto help you choose .
