Models
Google releases DiffusionGemma as open-weight model with 500+ tokens per second
Google has released DiffusionGemma, a 26B open-weight model using diffusion-based decoding, achieving over 500 tokens per second throughput on NVIDIA's NIM cloud API.
1 min read
SourceSimon Willison
Google released DiffusionGemma, a 26-billion-parameter open-weight model under Apache 2 license, bringing experimental diffusion-based text generation research into production. The model is now available on Hugging Face and hosted free on NVIDIA's NIM cloud API, marking a shift from Google's May 202...
Sign in to read the full analysis
Free account. Full analysis on LLM unit economics, plus the weekly Cost-of-Inference column.
Method & sources
- Source type
- Primary publication (lab/vendor blog) — our analysis + implication
- Source link
- Simon Willison
- Published
- UTC
- Byline
- By the gotcontext.ai team (editorial standards)
- Correction?
- corrections@gotcontext.ai