CancelledOpenAI
GPT-6.1 Astra
Planned GPT-6 frontier successor cancelled following internal safety reviews regarding unconstrained autonomous agent actions.
Date to be announced✓ Sourced
The release index
Reasoning AI releases
Follow models that focus on multi-step deliberation, mathematics and frontier scientific problem-solving.
Cancelled
ExpectedGoogle DeepMind
Gemini 4 Frontier
Next-generation frontier multimodal model family from Google DeepMind undergoing post-training scaling for 2027.
2027✓ Sourced
DelayedMeta
Llama 4 Behemoth
Massive 2-trillion parameter teacher model developed by Meta AI currently postponed for extended training and evaluation.
Date to be announced✓ Sourced
ExpectedxAI
Grok 5
Next-generation flagship reasoning and code model currently in training on xAI Colossus supercomputing clusters.
2027✓ Sourced
DelayedDeepSeek
DeepSeek-R2
Second-generation pure reinforcement learning reasoning model postponed for specialized domestic hardware adaptation.
Date to be announced✓ Sourced
ExpectedDeepSeek
DeepSeek V4-Pro Weights
Open-weight foundation model packaging in final preparations following initial API serving evaluation benchmarks.
2026 · Q4✓ Sourced
ConfirmedAlibaba Cloud
Qwen 4
Multi-trillion parameter foundation model confirmed by Alibaba Cloud at Apsara Conference for 2027 launch.
2027✓ Sourced
ExpectedMistral AI
Mistral Autumn Generation
Next-generation European foundation model family in final testing stages from Mistral AI for late 2026 release.
2026 · Q4✓ Sourced