Reactions
11 postsCH 
Chubby♨️@kimmonismusIf that’s true, crazy times lie ahead: GPT-6 at the end of the month or early August, and Fable 5.1 also on the verge of release. That would mean timelines are shortening even further, and if GPT-6 also features new pre-training, things will get really exciting!…
- likes: 1.4K
- replies: 59
- bookmarks: 184
- reposts: 54
CH Chubby♨️@kimmonismusGPT-6 will presumably be announced or released in around four weeks. New pre-training, much larger, significantly better. Mythos was essentially the origin story, the myth of AI, because it continued to prove one thing: much larger models create much more intelligence and…
- likes: 1.5K
- replies: 59
- bookmarks: 222
- reposts: 79
RN Ramez Naam@ramezBigger AI models are coming, both from the proprietary leaders and Chinese open weight vendors. https://twitter.com/synthwavedd/status/2074886230018568582
- likes: 16
- replies: 2
- bookmarks: 1
- reposts: 0
T( Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex> DeepSeek are preparing for V4 GA, which seems likely to be ≥ GLM-5.2, and have begun work on a new, larger model that will compete with the upcoming 2.7T MiniMax Pro sounds like pure wishcasting based on Minimax news but then again, why not. V4 is smol and undertrained…
- likes: 185
- replies: 8
- bookmarks: 23
- reposts: 6
T( Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞)@teortaxesTex1e25 is not a whole lot of compute. At most 1 month of 1 SuperPOD work (fp8). They have admitted that V4 is overcomplicated, they probably knew what how to make it better by the time it was completed. 100T tokens, AttnRes, repair the router etc But: doubt
- likes: 20
- replies: 0
- bookmarks: 2
- reposts: 1
LA Lisan al Gaib@scaling01OpenAI is back baby "GPT-6 will be based on a new, significantly larger pretrain than GPT-5.5/GPT-5.6" https://twitter.com/synthwavedd/status/2074886230018568582
- likes: 1.3K
- replies: 24
- bookmarks: 118
- reposts: 26