InayofuataMwongozo unaofuata
Jinsi ya Kuhimiza Miundo ya Kutoa Sababu
Lugha AI
MWONGOZO wa Makampuni
OpenAI o1 na o3 Miundo ya Kutoa Sababu Imefafanuliwa
OpenAI o1 na o3 ni miundo ya kutoa hoja inayotumia kokotoo la ziada la muda wa majaribio kutatua matatizo ya hatua nyingi katika hisabati, sayansi na usimbaji kabla ya kujibu.
Sikiliza mwongozo huuInaonyeshwa na AI · ~7 min
Katika ukurasa huudk 8 kusoma
Mambo muhimu ya kuchukua
- OpenAI o1, released in September 2024, was trained with reinforcement learning to produce a hidden chain of thought before answering.
- OpenAI reported o1 averaged 74% on AIME 2024 with one attempt, compared with 12% for GPT-4o.
- Reasoning models improve with more test-time compute, either by thinking longer per answer or by sampling many answers and choosing among them.
- o3 and o4-mini, released in April 2025, could use web search, Python and image tools inside their reasoning process.
- Hidden reasoning tokens are billed as output tokens, so reasoning models usually cost more and respond more slowly than chat models.
- OpenAI's PersonQA tests, reported by TechCrunch, found o3 hallucinated more often than o1, showing that longer reasoning does not guarantee accuracy.
Dive ya kina
OpenAI o1 and o3 are the first two generations of what OpenAI called its o-series, or reasoning models. Like earlier GPT models they are large language models: neural networks trained to predict the next token, a word or word fragment, of text. The difference is in what happens before the answer. A reasoning model first generates a long internal chain of thought, a running scratchpad of intermediate steps, checks and false starts, and only then writes its visible reply. OpenAI's announcement of o1 summarized the idea as a model that thinks before it answers. The thinking is produced as reasoning tokens, which OpenAI's API documentation says are not shown to developers but still take up room in the context window and are billed as output tokens.
The chain of thought is not just prompted; it is trained. OpenAI says o1 was trained with a large-scale reinforcement learning (RL) algorithm, a method in which a model tries many attempts and is rewarded for good outcomes rather than shown the correct text to copy. Through RL, according to the September 2024 announcement, the model learned to break hard steps into simpler ones, notice and correct its own mistakes, and switch strategies when an approach was failing. The company reported that o1's performance improved smoothly with more RL training compute and with more test-time compute, meaning more time spent thinking per question. That second axis was the headline: accuracy could be bought at inference time as well as by building a bigger model.
The timeline is short. OpenAI released o1-preview and o1-mini on 12 September 2024, followed by the full o1 in December 2024 alongside a system card describing its safety testing. On 20 December 2024 it previewed o3, and the ARC Prize Foundation published results for that preview on ARC-AGI-1, a puzzle benchmark designed to test novel pattern reasoning. o3-mini, a smaller and cheaper model aimed at science, math and coding, launched on 31 January 2025 and became the first reasoning model offered to free ChatGPT users. The full o3 and the smaller o4-mini arrived on 16 April 2025, and o3-pro followed for Pro subscribers in June 2025. OpenAI has since moved on: its posts now note that o1 is deprecated and that o3 and o4-mini are no longer available in ChatGPT and are deprecated in the API, with newer GPT-series reasoning models taking their place.
The published numbers explain the attention. On the 2024 AIME, a qualifying exam for the USA Mathematical Olympiad, OpenAI reported GPT-4o solved 12% of problems on average, while o1 averaged 74% with a single attempt, 83% with a majority vote over 64 samples, and 93% when re-ranking 1,000 samples with a learned scoring function. On Codeforces programming contests o1 reached the 89th percentile, and on GPQA Diamond, a set of graduate-level science questions, OpenAI said o1 was the first model to exceed the PhD experts it recruited for comparison. For o3, OpenAI reported 20% fewer major errors than o1 on difficult real-world tasks in external expert evaluations, and the ARC Prize Foundation reported the December 2024 preview scored 75.7% on its semi-private ARC-AGI-1 set at about $26 per task, and 87.5% using roughly 172 times more compute.
o3 also changed how reasoning connects to tools. OpenAI said o3 and o4-mini were trained with RL to decide when and how to use tools such as web search, a Python interpreter and image generation, and could chain several tool calls inside one reasoning process. They were also the first OpenAI models to put images into the chain of thought, cropping or rotating a picture mid-reasoning. With a Python interpreter, OpenAI reported o4-mini reached 99.5% on AIME 2025 on a single attempt, while cautioning that tool-assisted scores should not be compared with scores without tools.
Compared with a standard chat model, a reasoning model trades speed and cost for accuracy on problems with checkable answers. OpenAI's own human preference tests found o1-preview strongly preferred for data analysis, coding and math but not preferred on some natural language tasks such as writing. Compared with open alternatives, DeepSeek-R1, released in January 2025 and later published in Nature, showed that similar reasoning behavior, including self-reflection and verification, could emerge from reinforcement learning without human-written reasoning examples, and its authors described the training method publicly, whereas OpenAI has not published the o-series architecture or training recipe.
The limitations are well documented. The raw chain of thought is hidden: OpenAI chose to show users only a model-written summary, citing user experience, competitive advantage and the goal of monitoring the unaltered reasoning for problems such as manipulation. More thinking does not guarantee more truth. TechCrunch, citing OpenAI's o3 system card, reported that on OpenAI's PersonQA test of facts about people o3 hallucinated in 33% of answers and o4-mini in 48%, against 16% for o1, and the nonprofit lab Transluce found o3 sometimes claimed to have run code it could not have run. Benchmark context matters too: the ARC Prize Foundation noted that the o3 it tested had been trained on part of the ARC public training set and was not the same version later released.
In practice, reasoning models cost more per answer than chat models because every hidden reasoning token is billed. OpenAI's documentation recommends reserving at least 25,000 tokens for reasoning and output when starting out, warns that a response can run out of room before any visible text appears, and offers a reasoning effort setting and a cap on output tokens to control spending. Latency ranges from seconds to minutes. For simple lookups or casual writing, a faster non-reasoning model is usually the better fit.
03Mfano uliofanya kazi
Seeing what test-time compute buys on one problem set
Pick 15 problems with exact numeric answers, such as a past AIME exam, so every answer can be checked automatically.
Ask a reasoning model each question once at low reasoning effort and record accuracy, latency and the reasoning token count reported in the API usage field.
Repeat at high reasoning effort and compare: accuracy usually rises, and reasoning tokens, cost and latency rise with it.
For a parallel approach, sample each question 8 times at medium effort and take the most common answer as a majority vote.
Compute cost per correct answer for each setting by multiplying total tokens by the published price and dividing by the number correct.
Repeat the whole test with a non-reasoning chat model as a baseline.
Nini inaonyesha
A small table of accuracy, latency and cost per correct answer for each setting. It should show the pattern OpenAI reported for o1, accuracy rising with test-time compute, and make the price of that gain concrete.
Utekelezaji wa Ulimwengu Halisi
Solving AIME 2024 competition math: OpenAI reported o1 averaged 74% of problems with one sample, against 12% for GPT-4o, and 93% when re-ranking 1,000 samples.
Competitive programming: a model fine-tuned from o1 scored 213 points at the 2024 International Olympiad in Informatics under contest rules, placing in the 49th percentile.
Multi-step research in ChatGPT with o3, which OpenAI said could search the web, run Python on uploaded data and generate charts, chaining tool calls in a single answer.
Reading blurry whiteboard photos or hand-drawn diagrams with o3 and o4-mini, which OpenAI said could crop, zoom and rotate images as part of their reasoning.
Tuning cost and latency in the API with o3-mini's three reasoning effort settings, low, medium and high, so simple requests think briefly and hard ones think longer.
Mustakabali wa OpenAI o1 na o3 Miundo ya Kutoa Sababu Imefafanuliwa
OpenAI described o3 as pushing an additional order of magnitude of compute into both RL training and inference-time reasoning while still seeing gains, and it has since folded reasoning into newer GPT-series models rather than a separate o-series line. Several questions remain open. The hallucination results reported for o3 and o4-mini in April 2025 show that longer reasoning can raise confident errors on factual questions, and the cause is not settled. Chain-of-thought monitoring, which OpenAI cited as a reason to keep reasoning hidden but unaltered, depends on the chain staying faithful to the model's actual computation, which is an active research topic. Benchmarks are saturating: the ARC Prize Foundation called ARC-AGI-1 saturating in 2024 and launched the harder ARC-AGI-2, on which it said early o3 data pointed to under 30% even at high compute. Open-weight work that began with DeepSeek-R1 in January 2025 has made reasoning training reproducible outside OpenAI, so cost per correct answer, not only peak scores, is becoming the main point of comparison.
Kuiweka katika vitendo
Kwa nini ni muhimu
- Mkakati wa muuzaji. Ramani za barabara za wachuuzi huathiri vipengele ambavyo timu yako inaweza kuunda baadaye.
- Gharama na bajeti. Masharti ya kibiashara na chaguzi za kupeleka huathiri gharama na hatari ya muda mrefu.
- Hatari na usalama. Vivutio vya kampuni hutengeneza chaguo-msingi za bidhaa, mkao wa usalama na uwazi.
Jihadharini na
- Matangazo ya uzinduzi yanaweza kushinda uthabiti katika utendakazi halisi wa uzalishaji.
- Bei za API au mabadiliko ya sera yanaweza kuvunja mawazo mara moja.
- Utegemezi wa muuzaji mmoja huongeza gharama za kufunga na kuhama.
Hatua zinazofuata
- Tathmini watoa huduma kwa kutumia kazi na seti zako za data.
- Kagua faragha, usalama na masharti ya kisheria kabla ya kuunganishwa.
- Dumisha mpango mbadala kwa miundo au wachuuzi.
- Fuatilia maelezo ya toleo ili mabadiliko ya ramani ya barabara yasiwashangaze timu.
Vyanzo na kusoma zaidi
- OpenAILearning to reason with LLMs
- OpenAIIntroducing OpenAI o3 and o4-mini
- OpenAIOpenAI o3-mini
- OpenAI API documentationReasoning models
- arXiv (OpenAI)OpenAI o1 System Card
- ARC Prize FoundationOpenAI o3 Breakthrough High Score on ARC-AGI-Pub
- TechCrunchOpenAI's new reasoning AI models hallucinate more
- arXiv (published in Nature, 2025)DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
Endelea Kuchunguza
Jarida halali
Pata taarifa ya kila siku ya AI
Hadithi tatu za AI zilizothibitishwa kila asubuhi ya wiki, zimetengenezwa kwa Kiingereza rahisi. Bila malipo daima, hakuna matangazo.
Barua pepe moja kila siku ya wiki. Jiondoe kwa bonyeza moja. Hatutauza au kushiriki anwani yako.
Jijaribu mwenyewe
Take the OpenAI o1 and o3 Reasoning Models Explained quiz
Maoni ya haraka juu ya kila jibu, na cheti kinachoweza kushirikishwa na kitambulisho kinachoweza kuthibitishwa baada ya kupita kozi.
Changia elimu ya AI bila malipo. AI Understanding ni 501(c)(3) isiyo na faida — hakuna matangazo, hakuna kizuizi cha malipo, daima. Toa mchango
Maswali yanayoulizwa mara kwa mara
Je, OpenAI o1 na o3 Miundo ya Kutoa Sababu Inafafanuliwa nini?
OpenAI o1 na o3 ni miundo ya kutoa hoja inayotumia kokotoo la ziada la muda wa majaribio kutatua matatizo ya hatua nyingi katika hisabati, sayansi na usimbaji kabla ya kujibu.
What does OpenAI’s o-series documentation say the models use before returning a final answer?
OpenAI describes internal reasoning tokens, but this does not mean users receive a complete human-readable reasoning transcript.
What does OpenAI say is part of the o1 series training approach?
OpenAI describes reinforcement learning as part of o1 training for complex reasoning; avoid generalizing the specific training disclosure to every later model.
What does inference-time compute scaling mean for reasoning models?
OpenAI describes additional reasoning computation, but measured gains depend on the task and may increase token use or latency.
How should a reported benchmark score for an OpenAI reasoning model be interpreted?
Scores depend on model version, reasoning effort, benchmark setup, and evaluation harness; do not generalize one number to all tasks or versions.
Why might internal reasoning tokens affect API cost even when users do not see them?
OpenAI documentation explains that reasoning tokens consume output tokens and context capacity even when not shown in the final response.
Endelea kujifunza
Miongozo inayohusiana
Miongozo zaidi imechaguliwa kwa mada hii