Compare to
Discover how Open AI's GPT-4 and Google's Gemini Ultra stack up against each other in this comprehensive comparison of two leading AI
language models.
Released in March 2023 and December 2023 respectively, these models represent significant advancements in artificial intelligence,
with GPT-4 offering a 8,192-token context
window and Gemini Ultra featuring a 32,800-token
capacity. Their distinct approaches to natural language processing are reflected in their
benchmark performances, with GPT-4 achieving 86.4% on MMLU and Gemini Ultra scoring 83.7%, making this comparison essential
for developers and organizations seeking the right AI solution for their specific needs.
Models Overview
 GPT-4 |  Gemini Ultra | |
|---|---|---|
Provider Company that developed the model | Open AI | |
Context Length Maximum number of tokens the model can process | undefined | 32.8K |
Maximum Output Maximum number of tokens the model can generate in a single response | 8192 | 8192 |
Release Date Date when the model was released | 14-03-2023 | 06-12-2023 |
Knowledge Cutoff Training data cutoff date | September 2021 | Unknown |
Open Source Whether the model's code is open-source | FALSE | FALSE |
API Providers API providers that offer access to the model | OpenAI API | Vertex AI |
Pricing Comparison
Compare the pricing of Open AI's GPT-4 and Google's Gemini Ultra to determine the most cost-effective solution for your AI needs.
| GPT-4 | Gemini Ultra | |
|---|---|---|
Input Cost Cost per million input tokens | $30 / 1M tokens | Pricing not available |
Output Cost Cost per million tokens generated | $60 / 1M tokens | Pricing not available |
Comparing Benchmarks and Performance
Compare the performances of Open AI's GPT-4 and Google's Gemini Ultra on industry benchmarks. This section provides a detailed comparison on MMLU, MMMU, HumanEval, MATH and other key benchmarks.
| GPT-4 | Gemini Ultra | |
|---|---|---|
MMLU Evaluating LLM knowledge acquisition in zero-shot and few-shot settings. | 86.4% | 83.7% |
MMMU A wide ranging multi-discipline and multimodal benchmark. | 34.9% | 59.4% |
HellaSwag A challenging sentence completion benchmark. | 95.3% | Benchmark not available |
GSM8K Grade-school math problems benchmark. | 92% | 88.9% |
HumanEval A benchmark to measure functional correctness for synthesizing programs from docstrings. | 67% | 74.4% |
MATH Benchmark performance on Math problems ranging across 5 levels of difficulty and 7 sub-disciplines. | Benchmark not available | 53.2% |
