Philip Kiely's banner
Philip Kiely's profile picture

Philip Kiely

@philipkiely • 16,437 subscribers

Author of Inference Engineering | Building @baseten | Not an LLM (yet)

Shorts

Kimi K2 Thinking was already smarter than ChatGPT. Now, it is faster than ChatGPT.

Kimi K2 Thinking was already smarter than ChatGPT. Now, it is faster than ChatGPT.

100,331 просмотров

Videos

LIVE
1.2k

Anya Rossi

sweetdream.ai

SweetDream.aiSponsored•Livecam
Streaming Now

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

HD live stream
Exclusive private shows
1.2k viewers online

Free preview available • Premium content

Scaling past a trillion parameters requires inference engineers to solve 3 issues: - Width: Routing between hundreds of sparse experts - Depth: Mixing global and local linear attention - Length: Processing 1M-token sequences I asked Ali and Bryce to draw it up. They delivered.
1:01:44
philipkiely's profile picture

Scaling past a trillion parameters requires inference engineers to solve 3 issues: - Width: Routing between hundreds of sparse experts - Depth: Mixing global and local linear attention - Length: Processing 1M-token sequences I asked Ali and Bryce to draw it up. They delivered.

Philip Kiely

158,771 просмотров • 13 дней назад

On sample workloads: Opus 4.8 -> Kimi 2.7 Code | 82% savings GPT 5.5 -> GLM 5.2 | 77% savings Gemini 3.5 Flash -> Nemotron 3 Ultra | 67% savings Run the numbers for your yourself:
4:11
philipkiely's profile picture

On sample workloads: Opus 4.8 -> Kimi 2.7 Code | 82% savings GPT 5.5 -> GLM 5.2 | 77% savings Gemini 3.5 Flash -> Nemotron 3 Ultra | 67% savings Run the numbers for your yourself:

Philip Kiely

1,823,552 просмотров • 4 месяцев назад

Inference Engineering launches today.
1:11
philipkiely's profile picture

Inference Engineering launches today.

Philip Kiely

1,473,423 просмотров • 8 месяцев назад

The first 100 seconds of the Inference Engineering audiobook, narrated by a voice clone built by Rime. P.S. the video was also made using an AI-built script. Input is a PNG and a text file, script handles narration, bounding boxes, captions, and final video export.
1:40
philipkiely's profile picture

The first 100 seconds of the Inference Engineering audiobook, narrated by a voice clone built by Rime. P.S. the video was also made using an AI-built script. Input is a PNG and a text file, script handles narration, bounding boxes, captions, and final video export.

Philip Kiely

72,778 просмотров • 5 месяцев назад

Больше нет контента для загрузки