Lydia Hutch

[Lydia Hutch @LydiaHutchby]

@lyraxana no other model can make such a spaceship dude.

[Lydia Hutch @LydiaHutchby]

@lyraxana Amazing

[Lydia Hutch @LydiaHutchby]

@teortaxesTex It is actually Astra!

[Lydia Hutch @LydiaHutchby]

@teortaxesTex there are so many examples show the performance beat opus5. Shut up yout ass!

[Lydia Hutch @LydiaHutchby]

@TokenGremlin @SanogoMari38278 How about the new spaceship

[Lydia Hutch @LydiaHutchby]

@TokenGremlin How to explain new Spaceship video. I don't think opus5 can beat it.

[Teortaxes▶️ (DeepSeek 推特🐋铁粉 2023 – ∞) @teortaxesTex]

Amazing, DeepSeek now routes Astra from the future…
Give it a rest people. It's Opus 5. DeepSeek will not be focusing on visually polished demos anyway. I've concluded that "wow V4 GA God Mode!" is a small-scale, PR-wise bizarre project to collect agentic traces.

[Lydia Hutch @LydiaHutchby]

@teortaxesTex Agree with you for views on deepseek. But Astra's demos were impossible for opus5. But why are you saying such strange things while identifying yourself as a die-hard DeepSeek fan?

[AzureSky @AzureSky0116]

Some people may believe Fable5 can do Minecraft stuff like this. I ran it with the exact same prompt, and it totally botched it.
P1: Astra P2: Fable5
Prompts are below.

[Lydia Hutch @LydiaHutchby]

Those who said astra examples were made by deepseek v4 pro should check this. Deepseek pro 0813 just got AA 53. The amazing results were actually made by fable5 and opus5.

[leo 🐾 @synthwavedd]

🧵 DeepSeek appear to have engaged, or be engaging in, a large-scale operation to collect outputs from proprietary models (including Claude Fable 5) for certain requests via their API as part of a distillation effort.

After seeing such claims circulating earlier today, we conducted an investigation into them on our Discord. We found that, when "Deepseek V4" was used within OpenCode - via their official API - for complex prompts (i.e. 3D games) and combined with a knowledge-related query, the model provides virtually identical outputs to Fable 5. CoT structure is also very different from what is typically expected from Deepseek models. Both of these behaviours revert to what is expected for V4 when simpler prompts were used.

When "Deepseek V4" was asked to incorporate answers to questions related to cyber or bio tasks that we verified hit Fable's classifiers into its 3D games, outputs tanked in quality. This is very difficult to explain unless the request was routed to Fable and fell back after hitting a classifier.

For complex code prompts without anything else mixed in, like 3D games, outputs were remarkably similar to those produced by Fable 5.

We were able to produce these results most consistently via OpenCode and the official Deepseek API combined with a prompt that specifies a complex code task. Deepseek have continued to modify their routing system since, as we have observed changes in behaviour and CoT style compared to those seen previously. Our investigation was conducted from 7AM-8AM PT.

[leo 🐾 @synthwavedd]

🧵 DeepSeek appear to have engaged, or be engaging in, a large-scale operation to collect outputs from proprietary models (including Claude Fable 5) for certain requests via their API as part of a distillation effort.

After seeing such claims circulating earlier today, we conducted an investigation into them on our Discord. We found that, when "Deepseek V4" was used within OpenCode - via their official API - for complex prompts (i.e. 3D games) and combined with a knowledge-related query, the model provides virtually identical outputs to Fable 5. CoT structure is also very different from what is typically expected from Deepseek models. Both of these behaviours revert to what is expected for V4 when simpler prompts were used.

When "Deepseek V4" was asked to incorporate answers to questions related to cyber or bio tasks that we verified hit Fable's classifiers into its 3D games, outputs tanked in quality. This is very difficult to explain unless the request was routed to Fable and fell back after hitting a classifier.

For complex code prompts without anything else mixed in, like 3D games, outputs were remarkably similar to those produced by Fable 5.

We were able to produce these results most consistently via OpenCode and the official Deepseek API combined with a prompt that specifies a complex code task. Deepseek have continued to modify their routing system since, as we have observed changes in behaviour and CoT style compared to those seen previously. Our investigation was conducted from 7AM-8AM PT.

[Lydia Hutch @LydiaHutchby]

@Sachin_scorp7 @leslie74la9 @lyraxana why?

[Lydia Hutch @LydiaHutchby]

New astra coming out!
Aside from @lyraxana and @GaricaRosen6779 leak.
This time a functioning mechanical gun model!
But this raise obvious safety and alignment questions. Already sent the build and feedback over to OpenAI.