DeepSeek-R1
Incentivizing Reasoning Capability in LLMs via Reinforcement Learning
A language reasoning model release from DeepSeek, released with reinforcement-learning training and open model weights.
Pricing
Per 1M tokens; cache-miss input is USD 0.56, cache-hit input is USD 0.07, and output is USD 1.68.
DeepSeek's R1 launch documentation identifies the API model name as deepseek-reasoner. The current official pricing page lists deepseek-reasoner (currently DeepSeek-V3.1 Thinking Mode) at the stated per-1M-token rates, effective from 2025-09-05 16:00 UTC.
Model Intelligence
Recent stories
OpenAI says Jalapeño delivered 1.5–1.9× more work per watt and 1.7–3.6× lower end-to-end latency than NVIDIA systems in its tests. The company plans to deploy the inference chip in its compute infrastructure by year-end.
Kimi K3 posted strong coding results, including rank #5 on Artificial Analysis and #3 on DeepSWE. Engineers disputed whether its lower token price offsets higher token use and slower throughput.