Engineer Lists Inference Costs for Large AI Models
Research engineer Florian Brand lists inference prices for four large models of similar scale.
Florian Brand, a research engineer at Prime Intellect, replied with size and price figures for four models. Nemotron is listed at 550B parameters and 2.40/M. Trinity is listed at 398B and 0.80/M. Qwen MoE is listed at 397B and 3.50/M. GLM Flash is listed at 320B and 0.50/M. Brand added that readers should not focus on one parameter of a model. The post supplies no further verification or context beyond these stated values.
Combined views
25.8K
3 posts, first seen 2d ago
Engineer Lists Inference Costs for Large AI Models
Research engineer Florian Brand lists inference prices for four large models of similar scale.
Florian Brand, a research engineer at Prime Intellect, replied with size and price figures for four models. Nemotron is listed at 550B parameters and 2.40/M. Trinity is listed at 398B and 0.80/M. Qwen MoE is listed at 397B and 3.50/M. GLM Flash is listed at 320B and 0.50/M. Brand added that readers should not focus on one parameter of a model. The post supplies no further verification or context beyond these stated values.
Reactions
3 postsFB 
Florian Brand@xeophon@scaling01 Nemotron is 550B and costs 2.40/M Trinity is 398B and costs 0.80/M Qwen MoE is 397B and costs 3.50/M GLM Flash is 320B and costs 0.50/M Don’t focus on one param of a model
- likes: 48
- replies: 2
- bookmarks: 4
- reposts: 1
LA Lisan al Gaib@scaling01@xeophon bad comparison. not the pricing from one company within the same month
- likes: 16
- replies: 1
- bookmarks: 1
- reposts: 0
FB Florian Brand@xeophon@scaling01 i mean you could also look at the pricing of GLM-5.3 and 5.3 Flash to see that you can’t simply do "X times total params means X times output token pricing"
- likes: 4
- replies: 0
- bookmarks: 0
- reposts: 0