Kev-4B Decision Model Goes Live in Beta on Phala with NVIDIA Confidential Computing
bgmshana · x · 2026-10-14
Kev-4B, an open-weight System One decision model by Jared Palmer fine-tuned from Qwen3.5 4B Base, is now in beta on Phala Network inside NVIDIA TDX-attested enclaves. Instead of chat, it returns fast structured decisions — calibrated yes/no probabilities, option picks, or scale scores — via POST /v1/systemone with the TypeSafe SDK. Pricing is $0.04/M input tokens (output free) with an 8,192-token context.
Related event: Open-Source Decision Model Kev-4B Launches with TEE Private Inference(3 posts)→
More from Infra
- VirusTotal ships macOS Mirage sandbox to analyze fake password-prompt infostealers on Apple Silicon — jedisct1 · 2026-10-14
- ClickHouse merges IPCrypt support for GDPR-friendly IP pseudonymization — jedisct1 · 2026-10-14
- PS5LM on GitHub: llama.cpp port runs GGUF models natively on jailbroken PS5 — mhdfaran · 2026-10-14
- 4-bit keeps 98-99% quality, DeepSeek V4 Flash isn't a local-deployment model — gnukeith · 2026-10-14
- Fly.io Sprites adds connector for Ollama Cloud models — ollama · 2026-10-14
- CoreWeave faces $13.6B due by 2027 on $10B revenue as critics cry AI bubble — SumitGup · 2026-10-14