Models may hit 20T parameters this year, making inference the last moat
burkov · x · 2026-07-21
The post predicts that if open-weight models are already nearing 3T parameters and the rumored Fable model is 6–8T, then models above 20T parameters could appear by year-end.
- The author argues that inference for models this large may become the only remaining moat.
- Training corpora are also said to be measured in tens of trillions of tokens.
- The implication is roughly one floating-point number per token, which the author suggests may be enough to absorb the corpus without loss.
More from Infra
- 12 KV Cache Reduction Techniques Every AI Engineer Should Understand, Explained — blaizedsouza · 2026-09-11
- The shadow GPU capacity market is formalizing, with Meta selling excess compute to outside buyers — DavidLinthicum · 2026-09-11
- Engram's random reads don't suit SSDs; CPU-memory over NVLink could serve all 72 GPUs — bookwormengr · 2026-09-11
- 80% of the DIY LLM inference hype posters have already quit — it's brutally hard systems work — abhijithneil · 2026-09-11
- Hugging Face's Ultra Scale Playbook: a free book on training LLMs on GPU clusters — mdancho84 · 2026-09-11
- Is inference latency becoming the biggest bottleneck for production AI agents? — Euphoric_Sea632 · 2026-09-11