LLM Opex

Reports

LLM hosting prices: what changed in October 2026

196 price cuts and 99 rises; the largest is 100% on DeepInfra zai-org/GLM-5.3-Flash. 300 models newly sold per token, 225 models withdrawn, 124 retirements this month and 45 new GPU instances.

196 price cuts · 99 price rises · 300 models newly sold per token · 225 withdrawn · 45 new GPU instances · 126 new regions

Biggest price moves

CloudWhatPriceChange
DeepInfrazai-org/GLM-5.3-Flash, batch, global$0.012 → $0.024 +100%
DeepInfrazai-org/GLM-5.3-Flash, standard, global$0.015 → $0.03 +100%
DeepInfrazai-org/GLM-5.3-Flash, batch, global$0.20 → $0.40 per 1M output tokens+100%
DeepInfrazai-org/GLM-5.3-Flash, standard, global$0.25 → $0.50 per 1M output tokens+100%
DeepInfrazai-org/GLM-5.3-Flash, batch, global$0.06 → $0.12 per 1M input tokens+100%
DeepInfrazai-org/GLM-5.3-Flash, standard, global$0.075 → $0.15 per 1M input tokens+100%
Mistralmistral-large-4, global$2.09 → $4.18 per 1M output tokens+100%
Mistralmistral-large-4, global$0.07 → $0.14 +100%
Mistralmistral-large-4, global$0.68 → $1.36 per 1M input tokens+100%
DeepInfranvidia/NVIDIA-Nemotron-3.5-Super-VL, batch, global$0.064 → $0.016 -75%

And 285 more.

Models newly sold per token

Withdrawn and retiring

New GPU instances, by cloud

Alibaba Cloud 37, Core42 4, Azure 4

New regions, by cloud

Azure 92, AWS 31, Nebius 2, Fireworks AI 1

The clouds' own announcements

The month before · The month after