FRAMEWIREIndonesiaUpdated Aug 22Live wire
0:00 / 0:00

Top AI LLMs for Every Task

Try Gemini 3.1 Pro FREE on GlobalGPT 👉 → Writing & Research: GPT-5.4, Claude 4.6, Gemini 3.1 Pro, Perplexity → Social Content: Grok 4, GPT o3, DeepSeek → Academic / STEM: Claude Opus 4.6, MiniMax M2.7

LeonardoAug 221
0:00 / 0:00

DeepSeek V4 Pro is already huge

1.6T total parameters. 49B active per token. 1M-token context. And it uses token-wise compression + sparse attention so it doesn’t need to reread everything constantly. But long agent tasks can still drift.

Julian Goldie SEOAug 221
0:00 / 0:00

Deepseek releases experimental multi-modal model, what is the actual measured effect?

I used the exact same prompt, requiring deepseek-v4-flash-vision-exp and gemini-3.7-flash to use Three.js to independently build a "Deep Sea Bioluminescence Observation Cabin" with complete luminous particles and material effects. See the video for the final result. It is worth noting that compared to gemini-3.7-flash, DeepSeek…

洞寓法师🦁Aug 222
0:00 / 0:00

DeepSeek V4's biggest problem was never the model.

It was the harness. A GitHub dev built JSpace, a wrapper that changes ZERO weights. It just manages how the model works. Same model. Same brain. Scores jumped across every benchmark. → DeepSWE: 62.7 to 72.0 → NL2Repo: 61.5

Julian Goldie SEOAug 22