ModelClock
OFFICIAL NEBIUS NOTICES

Nebius model deprecation schedule

Nebius has no upcoming shutdown dates published. In the last 12 months 21 models were retired or shut down, most recently Qwen/Qwen3-Next-80B-A3B-Thinking on Aug 31, 2026.

Last updated · Nebius dashboard and alerts

Upcoming shutdowns 0

Nebius has not published an upcoming shutdown date for any tracked model.

Retired or shut down in the last 12 months 21

Nebius models retired in the last 12 months
DateModelStatusReplacementSource
Qwen/Qwen3-Next-80B-A3B-ThinkingShut downQwen/Qwen3.5-397B-A17BNotice ↗
Qwen/Qwen3-32BShut downnvidia/Nemotron-35-LightningNotice ↗
Qwen/Qwen2.5-VL-72B-InstructShut downMiniMaxAI/MiniMax-M3Notice ↗
nvidia/Nemotron-3-Nano-OmniShut downnvidia/Nemotron-35-LightningNotice ↗
nvidia/Llama-3_1-Nemotron-Ultra-253B-v1Shut downnvidia/nemotron-3-super-120b-a12bNotice ↗
nvidia/Cosmos3-Super-ReasonerShut downMiniMaxAI/MiniMax-M3Notice ↗
NousResearch/Hermes-4-70BShut downnvidia/Nemotron-35-LightningNotice ↗
MiniMaxAI/MiniMax-M2.5Shut downMiniMaxAI/MiniMax-M3Notice ↗
meta-llama/Llama-3.3-70B-InstructShut downnvidia/Nemotron-35-LightningNotice ↗
deepseek-ai/DeepSeek-V4-FlashShut downdeepseek-ai/DeepSeek-V4-Flash-0731Notice ↗
zai-org/GLM-5Shut downNone publishedNotice ↗
Qwen/Qwen3.5-397B-A17B-fastShut downNone publishedNotice ↗
Qwen/Qwen3-Next-80B-A3B-Thinking-fastShut downNone publishedNotice ↗
Qwen/Qwen3-235B-A22B-Thinking-2507-fastShut downNone publishedNotice ↗
PrimeIntellect/INTELLECT-3Shut downNone publishedNotice ↗
openai/gpt-oss-120b-fastShut downNone publishedNotice ↗
moonshotai/Kimi-K2.5-fastShut downNone publishedNotice ↗
moonshotai/Kimi-K2.5Shut downNone publishedNotice ↗
MiniMaxAI/MiniMax-M2.5-fastShut downNone publishedNotice ↗
deepseek-ai/DeepSeek-V3.2-fastShut downNone publishedNotice ↗
deepseek-ai/DeepSeek-V3.2Shut downNone publishedNotice ↗

Official Nebius sources

Each provider and hosting route sets its own dates. A model offered on several platforms can stop on different days; ModelClock reads each notice from the provider's own documentation.