Acellera burns 1B tokens a day: testing 7 local LLM setups on a single RTX 5090 for drug discovery
gdefabritiis · x · 2026-09-10
Acellera uses about 1B tokens per day (including input and cached tokens), so the author set out to see how far local models can go. They benchmarked 7 LLM + harness combinations on an IL-23 target discovery task, with several surprises — including what a single RTX 5090 could produce: one report slide deck was generated locally with Qwen3 27B plus PlayMolecule AI. Blog and docs are available.
Related event: Acellera Tests Local LLM Drug Discovery on a Single RTX 5090(2 posts)→
More from Infra
- 12 Core Microservices Communication Patterns Explained in One Visual — goyalshaliniuk · 2026-09-10
- Powering AI is an architecture problem, not a power problem, says MIT Tech Review — nordicinst · 2026-09-10
- Pocket AI Lab: open-source iOS app runs LLMs fully on-device with three backends — Ammoryyy · 2026-09-10
- AI data centers are an architecture problem: 3GW dropped in seconds — MIT Tech Review AI · 2026-09-10
- Triton creator Phil Tillet on Gluon: handing GPU decisions back to AI models — TheTuringPost · 2026-09-10
- DeepSeek open-sources deepseek-recipe to ease deploying V4.1 Flash and future models — teortaxesTex · 2026-09-10