DeepSeek-R1

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.

Command A

Command R+ is Cohere’s cutting-edge generative AI model, engineered for enterprise-grade performance where speed, security, and output quality are critical. Designed to run efficiently with minimal infrastructure, it outperforms top-tier models like GPT-4o and DeepSeek-V3 in both capability and cost-effectiveness. Featuring an extended 256K token context window—twice as large as most leading models—it excels at complex multilingual and agent-based tasks essential for modern business operations. Despite its power, it can be deployed on just two GPUs, making it highly accessible. With blazing-fast throughput of up to 156 tokens per second—about 1.75x faster than GPT-4o—Command R+ delivers exceptional efficiency without compromising accuracy or depth.

DeepSeek-R1Command A
Provider
Web Site
Release Date
Jan 21, 2025
3 months ago
Mar 14, 2025
1 month ago
Modalities
text ?
text ?
API Providers
DeepSeek, HuggingFace
Cohere, Hugging Face, Major cloud providers
Knowledge Cut-off Date
Unknown
-
Open Source
Yes
Yes
Pricing Input
$0.55 per million tokens
$2.50 per million tokens
Pricing Output
$2.19 per million tokens
$10.00 per million tokens
MMLU
90.8%
Pass@1
Source
85.5%
Source
MMLU Pro
84%
EM
Source
Not available
MMMU
-
Not available
HellaSwag
-
Not available
HumanEval
-
Not available
MATH
-
80%
Source
GPQA
71.5%
Pass@1
Source
50.8%
Source
IFEval
83.3%
Prompt Strict
Source
90.9%
Source
Mobile Application
-

Compare LLMs

Add a Comment


10%
Our site uses cookies.

Privacy and Cookie Policy: This site uses cookies. By continuing to use the site, you agree to their use.