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.
Grok 3 is xAI's most advanced model, trained on the Colossus supercluster with 10 times the computational power of previous state-of-the-art models. It boasts a 1M-token context window and advanced reasoning capabilities, enhanced through large-scale reinforcement learning, enabling deep thought processes ranging from seconds to minutes for solving complex problems. The model achieves top-tier performance across academic benchmarks and real-world user evaluations, earning an Elo score of 1402 in the Chatbot Arena. It was released alongside Grok 3 Mini, a cost-efficient variant optimized for streamlined reasoning.
| DeepSeek-R1 | Grok 3 Beta | |
|---|---|---|
Provider | ||
Web Site | ||
Release Date | Jan 21, 2025 3 months ago | Jan 19, 2025 3 months ago |
Modalities | text | text images video |
API Providers | DeepSeek, HuggingFace | xAI |
Knowledge Cut-off Date | Unknown | 2025-01 |
Open Source | Yes | No |
Pricing Input | $0.55 per million tokens | Not available |
Pricing Output | $2.19 per million tokens | Not available |
MMLU | 90.8% Pass@1 Source | Not available |
MMLU Pro | 84% EM Source | 79.9% Base model Source |
MMMU | - | 78% With Think mode Source |
HellaSwag | - | Not available |
HumanEval | - | Not available |
MATH | - | Not available |
GPQA | 71.5% Pass@1 Source | 84.6% With Think mode, Diamond Source |
IFEval | 83.3% Prompt Strict Source | Not available |
Mobile Application |
Compare AI. Test. Benchmarks. Mobile Apps Chatbots, Sketch
Copyright © 2025 All Right Reserved.