Deepseek R1
FreemiumFree no-login web interface and browser WebGPU access to the open-source DeepSeek R1 advanced reasoning AI model.
Informations sur l’outil
Qu'est-ce que Deepseek R1 ?
Deepseek R1 is a free no-login web interface and browser-based WebGPU access point for the open-source DeepSeek R1 advanced reasoning AI model. It offers free online chat and local in-browser inference for this MoE-based large language model focused on advanced reasoning, mathematics, and coding tasks. The site provides access to MIT-licensed model weights and an OpenAI-compatible API endpoint. Note that this is a fan-made website not affiliated with DeepSeek.
Fonctionnalités principales
MoE Architecture with Reinforcement Learning
Uses Mixture of Experts architecture with 37B active parameters, 671B total parameters, and 128K context length, incorporating advanced reinforcement learning for self-verification, multi-step reflection, and human-aligned reasoning.
Benchmark Performance
Reports 97.3% accuracy on MATH-500, 79.8% pass rate on AIME 2024, and performance exceeding 96.3% of Codeforces participants.
Free Online Chat
Provides DeepSeek R1 chat online free without requiring login.
WebGPU Local Browser Inference
Runs DeepSeek-R1-Distill-Qwen-1.5B entirely in the browser with WebGPU acceleration via Transformers.js and ONNX Runtime Web, supporting offline use after loading with no data sent to a server.
Open-Source Model Weights and Distillation
MIT-licensed weights available with base R1-Zero, enhanced R1, and distilled variants from 1.5B to 70B parameters for commercial use, optimized for complex problem-solving, multilingual understanding, and code generation.
OpenAI-Compatible API Endpoint
Offers an OpenAI-compatible API endpoint.
Cas d'utilisation
- #1
Advanced Reasoning Tasks
Supports complex multi-step reasoning, self-verification, and human-aligned problem solving.
- #2
Mathematical Reasoning
Handles advanced mathematics with high reported accuracy on MATH-500 and AIME benchmarks.
- #3
Code Generation and Programming
Enables production-grade code generation and competitive programming performance.
- #4
Local In-Browser Inference
Allows running distilled reasoning models offline in the browser without server communication.
Plans tarifaires
Free online chat with no login required; free open-source MIT-licensed model weights and distilled variants for commercial use; OpenAI-compatible API endpoint listed at $0.14 per million tokens; free account signup option noted.
Alternatives
Foire aux questions
Is Deepseek R1 free to use?
Can I run the model offline in my browser?
Is this the official DeepSeek website?
What architecture does DeepSeek R1 use?
What benchmarks does the model report?
Are the model weights open source?
Deepseek R1 Contacts et réseaux sociaux
- Assistance
- Site Internet :
- GitHub
Avis
Partagez votre expérience avec cet outil...