Recommended Free Tools
OpenAI announced GPT-4 on March 14, 2023, describing it as a multimodal AI model that accepts text and images and produces text. The headline improvement over GPT-3.5 was stronger performance on complex tasks, but OpenAI also warned that GPT-4 can invent facts and make reasoning errors. At launch, image input was not generally available, and the announcement’s access details are historical rather than current guidance.
What OpenAI announced
OpenAI’s March 14, 2023 announcement called GPT-4 a large multimodal model: it can take text and image inputs and return text. OpenAI characterized its results as human-level on various professional and academic benchmarks, while noting that the model remained less capable than people in many real-world situations. Benchmark performance is evidence about particular tests, not proof of general human-level competence or a guarantee of performance at work.
The announcement covered the model’s capabilities and limitations, rather than a fully disclosed technical design. OpenAI’s 2023 technical report says GPT-4 was pretrained to predict the next token using publicly available data, including internet data, and data licensed from third-party providers. It was then fine-tuned with reinforcement learning from human feedback (RLHF).
The report does not disclose the model’s size, hardware, training compute, dataset construction, or comparable implementation details. OpenAI cited competitive and safety considerations as its reasons for withholding those details.
#1 Best Overall
How GPT-4 differed from GPT-3.5
OpenAI said the difference could be hard to notice in casual conversation but became clearer when a task was sufficiently complex. It described GPT-4 as more reliable, creative, and better at following nuanced instructions than GPT-3.5. Those are OpenAI’s comparative claims, not a promise that GPT-4 will outperform GPT-3.5 on every prompt.
| Comparison | What OpenAI said at launch | Important qualification |
|---|---|---|
| Task difficulty | GPT-4’s advantage became more apparent as task complexity increased. | Differences in ordinary conversation could be subtle. |
| Input and output | GPT-4 was described as accepting text and images and producing text. | Image input was in limited alpha or research preview at launch, not broadly available. |
| Exam performance | GPT-4 scored around the top 10% of simulated bar-exam test takers; GPT-3.5 scored around the bottom 10%. | This was an OpenAI-reported simulated-exam result, not a measure of professional competence. |
| Reliability | OpenAI reported improvements in factuality and responses to disallowed-content requests in internal evaluations. | These internal comparisons do not guarantee factual or safe responses in an individual use. |
What the bar-exam result does—and does not—show
OpenAI reported that GPT-4 performed around the top 10% of simulated bar-exam test takers, compared with GPT-3.5 around the bottom 10%. The figure refers to a simulated exam and should not be read as evidence that GPT-4 can practice law, provide dependable legal advice, or perform like a lawyer in real cases.
OpenAI said it used recent publicly available tests or purchased 2022–2023 practice-exam editions and did not specifically train GPT-4 for those exams. It also acknowledged that a minority of exam problems had been seen during training. The result is a notable benchmark comparison, but it does not establish how the model performs across the full range of legal work.
Image input and launch-era access
At launch, OpenAI said text capability was being released through ChatGPT and its API. ChatGPT Plus access was subject to a usage cap, while API access was via a waitlist. Image input remained a limited alpha or research preview. These details describe the March 2023 announcement, not current access, pricing, model names, or terms.
Free tools Windows power users keep installed
One-click scans. No signup required.
OpenAI later announced API general availability for eligible developers. Its API availability page, published in 2023 and updated April 24, 2024, said existing API developers with a history of successful payments could access GPT-4 with 8K context at that time. That dated announcement does not establish present-day availability or terms.
Safety improvements and remaining limitations
OpenAI reported six months of iterative alignment work, including adversarial testing and feedback. In its March 2023 release overview, it said GPT-4 was 82% less likely than GPT-3.5 to respond to disallowed-content requests and 40% more likely to produce factual responses in internal evaluations. These are OpenAI’s evaluation figures, not independent results or guarantees.
OpenAI’s central warning was that GPT-4 is not fully reliable: it can hallucinate facts and make reasoning errors. Its release materials also identify social biases and adversarial prompts as limitations. The technical report discusses broader risks including disinformation, over-reliance, privacy, cybersecurity, and proliferation. Human review, grounding answers in relevant material, and avoiding high-stakes use can reduce some risks, but do not eliminate them.
OpenAI also released Evals, an open-source framework intended to help developers create benchmarks, inspect performance sample by sample, identify shortcomings, and catch regressions. Evaluation can reveal problems; it does not make a model’s outputs inherently reliable.
What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
What remains unclear
OpenAI’s launch materials support a specific conclusion: GPT-4 was presented as a more capable successor to GPT-3.5 on demanding tasks, with image understanding among its described capabilities and significant reliability caveats. They do not establish broad professional equivalence, disclose many technical details, or confirm current product access and pricing. Those distinctions matter when interpreting the 2023 announcement as well as when deciding how much to trust a model’s output.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
