Announcing our partnership with OpenAI. Read more

Ali Taha

Ali Taha

Foundations

22,580: GPT-2 to Kimi K3, explained

Ali Taha

Ali Taha

Twenty-two thousand five hundred and eighty. That’s how many GPT-2 (2019) models fit inside KimiK3 (2026).

Model performance

Real-time video generation inference on Baseten

Ali Taha

Brendan Duke

Yikai Zhu

Faraz Shahsavan

Pankaj Gupta

Philip Kiely

Ali Taha

5 others

Real-time video generation inference

Model performance

Sub-second image generation with Flux.2 and Qwen-Image

Faraz Shahsavan

Yikai Zhu

Ali Taha

Pankaj Gupta

Faraz Shahsavan

3 others

Sub-second image generation with Flux.2 and Qwen-Image

Model performance

I spent 31 hours on the math behind TurboQuant so you don't have to

Ali Taha

Ali Taha

Turboquant

Model performance

Four Bits: 4-Bit Quantization for Inference Optimization in FLUX.2 [dev]

Ali Taha

Ali Taha

four bits

Explore Baseten today

Start deploying Talk to an engineer