This should not be this cheap
Moon dev: "everyone is paying insane money for AI when there's one that costs almost nothing and somehow beats it" "i put deepseek head to head with something called GLM flash and what happened flipped everything i thought i knew"
Verdent AI just killed the barrier to software development.
No more expensive coding paywalls. No more getting stuck on complex bugs. Here’s the Verdent AI play 👇 → Toggle 3 reasoning levels for custom AI logic. → Turn website screenshots into functional code. → Use GLM
DeepSeek-V4-Pro-0813 is live on CoreWeave Serverless Inference.
1.6T parameters. 1M context. Built for long-horizon agent work, and priced for it: cache reads run $0.044/M for all the context your agent re-sends every step. Start here:
DeepSeek put out a vision version of V4 Flash and the repo weighs 305B parameters.
The text-only one is 304B. So vision cost them a billion parameters, which would be remarkable if that were the whole story. A month ago I went through the text repo. The model itself is 284B. The
You have reached the end of the archive
All of deepseek