Yum⋆₊˚ (@yuhasbeentaken)
@yuhasbeentaken
Growth engineer | I benchmark AI agents & build @offloopHQ, multi-agent multi-player workspace we’re hiring, DM open| prev @Stanford @tiktok_us @weixin_wechat
750 正在关注 3.7K 粉丝
deepseek v4 is terrible at hallucinations...
>deepseek v4 pro scores 94%.
>deepseek v4 flash is even worse at 96%.
for comparison:
>glm-5.2: 28%
>mimo v2.5 pro: 25%
>minimax m3: 16%
lower is better.
this means deepseek often answers confidently even when it should admit it doesn’t know.
use glm-5.2 for execution...
but never trust the output without tests, verification or a second model reviewing the work!!!
显示更多
deepseek v4 is coming in 2 weeks...
5 things we know:
1. v4 is expected to launch officially in mid-july
2. there will be pro and flash versions
3. deepseek is introducing peak and off-peak api pricing
4. prices will double during 7 peak hours per day
5. deepseek promises feature optimizations and performance improvements
5 things that are still speculation:
1. native vision and multimodal input
2. a new checkpoint rather than the current preview model
3. engram memory being included
4. a major intelligence jump toward glm-5.2 or kimi k2.7
5. the price increase being caused by new hardware and backend infrastructure
the most likely outcome?
a more stable, faster and production-ready version of v4 preview...
possibly with new capabilities unlocked, but not necessarily a dramatically smarter model.
显示更多
deepseek v4 is coming in 2 weeks...
5 things we know:
1. v4 is expected to launch officially in mid-july
2. there will be pro and flash versions
3. deepseek is introducing peak and off-peak api pricing
4. prices will double during 7 peak hours per day
5. deepseek promises feature optimizations and performance improvements
5 things that are still speculation:
1. native vision and multimodal input
2. a new checkpoint rather than the current preview model
3. engram memory being included
4. a major intelligence jump toward glm-5.2 or kimi k2.7
5. the price increase being caused by new hardware and backend infrastructure
the most likely outcome?
a more stable, faster and production-ready version of v4 preview...
possibly with new capabilities unlocked, but not necessarily a dramatically smarter model.
显示更多
here’s the list of western companies moving ai workloads to chinese models:
1. lindy → deepseek v4
2. cursor → kimi k2.5
3. coinbase → glm-5.2 + kimi 2.7
4. shopify → qwen
5. airbnb → qwen
6. uber eats → qwen2
7. siemens → deepseek + qwen
8. chapsvision → qwen
9. microsoft → testing deepseek v4
it’s becoming a procurement story!!!
显示更多
