LLM hosting prices: what changed in October 2026
Month in progress, updated 5 October 2026.
15 price cuts and four rises so far; the largest is 20% off Azure Standard_NC4as_T4_v3. One model newly sold per token, one model withdrawn, 28 retirements this month and 191 new GPU instances.
15 price cuts · 4 price rises · 1 models newly sold per token · 1 withdrawn · 191 new GPU instances · 46 new regions
Biggest price moves
| Cloud | What | Price | Change |
|---|---|---|---|
| Azure | Standard_NC4as_T4_v3, westeurope | $0.3657 → $0.2925 an hour | -20% |
| Azure | Standard_NC16as_T4_v3, westeurope | $0.8365 → $0.6691 an hour | -20% |
| Azure | Standard_NC8as_T4_v3, westeurope | $0.5225 → $0.4179 an hour | -20% |
| Azure | Standard_NC64as_T4_v3, westeurope | $3.02 → $2.42 an hour | -20% |
| Azure | Standard_NC40ads_H100_v5, westeurope | $4.18 → $3.35 an hour | -20% |
| Azure | Standard_NC80adis_H100_v5, westeurope | $8.37 → $6.69 an hour | -20% |
| Google Cloud | a3-highgpu-8g, spot, us-central1 | $52.96 → $50.32 an hour | -5% |
| Google Cloud | a3-megagpu-8g, spot, us-central1 | $55.89 → $53.10 an hour | -5% |
| Google Cloud | g2-standard-4, spot, us-central1 | $0.4241 → $0.4028 an hour | -5% |
| Google Cloud | g2-standard-8, spot, us-central1 | $0.5121 → $0.4865 an hour | -5% |
And 9 more.
Models newly sold per token
- Core42: gpt-6-sol from $2.20 in / $11.00 out per 1M tokens
Withdrawn and retiring
- Core42 no longer sells k2-horizon-375b
- AWS retires Claude Sonnet 4 on 2026-10-14
- Azure retires gpt-4.1-nano (2025-04-14) on 2026-10-14
- Azure retires gpt-4o-mini-transcribe (2025-03-20) on 2026-10-15
- Azure retires gpt-4o-mini-tts (2025-03-20) on 2026-10-15
- Azure retires gpt-4o-transcribe (2025-03-20) on 2026-10-15
- Azure retires sora-2 (2025-12-08) on 2026-10-15
- Azure retires Cohere-command-a-plus-05-2026 (1) on 2026-10-16; replacement: coherelabs-command-a-plus-05-2026-fp81, coherelabs-command-a-plus-05-2026-bf161, coherelabs-command-a-plus-05-2026-w4a41
- Google Cloud retires gemini-2.5-flash on 2026-10-20; replacement: gemini-3.8-flash or gemini-3.5-flash-lite or gemini-3.1-flash-lite
- Google Cloud retires gemini-2.5-flash-lite on 2026-10-20; replacement: gemini-3.8-flash or gemini-3.1-flash-lite or Gemma 4
- Google Cloud retires gemini-2.5-pro on 2026-10-20; replacement: gemini-3.8-flash or gemini-3.5-flash
- Azure retires gpt-image-1 (2025-04-15) on 2026-10-23
- OpenAI retires ft-babbage-002 on 2026-10-23; replacement: gpt-5.6-terra
- OpenAI retires ft-davinci-002 on 2026-10-23; replacement: gpt-5.6-terra
- OpenAI retires ft-gpt-3.5-turbo on 2026-10-23; replacement: gpt-5.6-terra
- OpenAI retires ft-gpt-4 on 2026-10-23; replacement: gpt-5.6-sol
- OpenAI retires ft-gpt-4.1-nano-2025-04-14 on 2026-10-23; replacement: gpt-5.6-luna
- OpenAI retires ft-o4-mini-2025-04-16 on 2026-10-23; replacement: gpt-5.6-terra
- OpenAI retires gpt-3.5-turbo-0125 on 2026-10-23; replacement: gpt-5.6-terra
- OpenAI retires gpt-4-0613 on 2026-10-23; replacement: gpt-5.6-sol
- OpenAI retires gpt-4-1106-preview on 2026-10-23; replacement: gpt-5.6-sol
- OpenAI retires gpt-4-turbo on 2026-10-23; replacement: gpt-5.6-sol
- OpenAI retires gpt-4.1-nano on 2026-10-23; replacement: gpt-5.6-luna
- OpenAI retires gpt-4o-2024-05-13 on 2026-10-23; replacement: gpt-5.6-sol
- OpenAI retires gpt-image-1 on 2026-10-23; replacement: gpt-image-2.5-sunburst or gpt-image-2.5-flare
- OpenAI retires o1-2024-12-17 on 2026-10-23; replacement: gpt-5.6-sol
- OpenAI retires o1-pro-2025-03-19 on 2026-10-23; replacement: gpt-5.6-sol (reasoning.mode: pro)
- OpenAI retires o3-mini-2025-01-31 on 2026-10-23; replacement: gpt-5.6-sol
- OpenAI retires o4-mini-2025-04-16 on 2026-10-23; replacement: gpt-5.6-terra
New GPU instances, by cloud
Alibaba Cloud 180, Azure 11
New regions, by cloud
AWS 25, Azure 21