DeepSeek-R1 is a 671B parameter Mixture-of-Experts (MoE) model with 37B activated parameters per token, trained via large-scale reinforcement learning with a focus on reasoning capabilities. It incorporates two RL stages for discovering improved reasoning patterns and aligning with human preferences, along with two SFT stages for seeding reasoning and non-reasoning capabilities. The model achieves performance comparable to OpenAI-o1 across math, code, and reasoning tasks.
GPT-4.1, launched by OpenAI on April 14, 2025, introduces a 1 million token context window and supports outputs of up to 32,768 tokens per request. It delivers outstanding performance on coding tasks, achieving 54.6% on the SWE-Bench Verified benchmark, and shows a 10.5% improvement over GPT-4o on MultiChallenge for instruction following. The model's knowledge cutoff is set at June 2024. Pricing is $2.00 per million tokens for input and $8.00 per million tokens for output, with a 75% discount applied to cached inputs, making it highly cost-efficient for repeated queries.
| DeepSeek-R1 | GPT-4.1 | |
|---|---|---|
Web Site
| ||
Provider
| ||
Chat
| ||
Release Date
| ||
Modalities
| text | text images |
API Providers
| DeepSeek, HuggingFace | OpenAI API |
Knowledge Cut-off Date
| Unknown | - |
Open Source
| Yes | No |
Pricing Input
| $0.55 per million tokens | $2.00 per million tokens |
Pricing Output
| $2.19 per million tokens | $8.00 per million tokens |
MMLU
| 90.8% Pass@1 Source | 90.2% pass@1 Source |
MMLU-Pro
| 84% EM Source | - |
MMMU
| - | 74.8% Source |
HellaSwag
| - | - |
HumanEval
| - | - |
MATH
| - | - |
GPQA
| 71.5% Pass@1 Source | 66.3% Diamond Source |
IFEval
| 83.3% Prompt Strict Source | - |
SimpleQA
| - | - |
AIME 2024 | - | 48.1% Source |
AIME 2025 | - | - |
Aider Polyglot
| - | - |
LiveCodeBench v5
| - | - |
Global MMLU (Lite)
| - | 87.3% pass@1 Source |
MathVista
| - | - |
Mobile Application |
Compare AI. Test. Benchmarks. Mobile Apps Chatbots, Sketch
Copyright © 2025 All Right Reserved.