←back to thread

Smollm3: Smol, multilingual, long-context reasoner LLM

(huggingface.co)
347 points kashifr | 2 comments | 08 Jul 25 16:13 UTC | HN request time: 0.395s | source
Show context
WhitneyLand ◴[08 Jul 25 17:27 UTC] No.44502146[source]▶
Mostly SOTA performance at the 3B level. A notable addition to the small but truly open club of models that provide full disclosure, code, recipes to reproduce their work.

Looks like ballpark a million dollars of GPU time if you want to train up one for yourself (4000 gpus/24 days).

Very nice write up that’s generous in sharing their learnings.

This is a solid and positive contribution.

replies(2): >>44502692 #>>44504060 #
YetAnotherNick ◴[08 Jul 25 18:31 UTC] No.44502692[source]▶
It's 384 H100s for 24 days, costing less than half a million dollars.
replies(2): >>44503252 #>>44505653 #
segmondy ◴[09 Jul 25 01:53 UTC] No.44505653[source]▶
H100 are going for about $3/hr, 384243 ~ $28k
replies(6): >>44505754 #>>44505979 #>>44506134 #>>44507506 #>>44507964 #>>44509849 #
1. jrk ◴[09 Jul 25 03:38 UTC] No.44506134[source]▶
This is indeed a reasonable cost estimate for competitive short-term H100 rentals (source: much SemiAnalysis coverage, and my own exploration of the market), but there is a critical error (besides the formatting glitch with `*`):

It was 24 days (576 hours) not 24 hours. $663,552 @ $3/hr.

replies(1): >>44509470 #
2. mromanuk ◴[09 Jul 25 12:54 UTC] No.44509470[source]▶
According to Runpod pricing page, you can run H100 for $2.39, it can go as lower as $528,629.76

WARNING: This is highly speculative and napkin math

H200 (141 GB HBM3 - $3.99/h - 1.4x perf) 216 x 24 x 17 = 88128h = 351.895,104 (17 days and 216 cards)

B200 (192 GB HBM3e - $5.99/h - 2.8x perf) 158 x 24 x 9 = 34128h = $204.426,72

Probably wrong math, should be more efficient and cheaper. Doubt that they have 100/200 cards available for that long.

Source: I've only trained using RTX4090 and stuff like that with 8 cards.

Not affiliated in any way with Runpod.