DeepSeek has cut API prices by up to 90% following its V4 model release, bringing cached input costs to RMB 0.025 per mi...

DeepSeek has cut API prices by up to 90% following its V4 model release, bringing cached input costs to RMB 0.025 per million tokens. The company says the cuts set a new industry low and are driven by its sparse attention architecture, which reduces per-token compute requirements to 27% of its predecessor. Context windows reach 1 million tokens. https://pandaily.com/deep-seek-slashes-api-prices-sets-new-low-for-large-model-costs #China #Tech #AI