smollm3-3b

A Core AI bundle of HuggingFaceTB/SmolLM3-3B for the Aether SDK (iOS and macOS 27+).

• Source: HuggingFaceTB/SmolLM3-3B at revision a07cc9a04f16550a088caea529712d1d335b0ac1, licence Apache-2.0. The licence is included as LICENSE. HuggingFaceTB/SmolLM3-3B declares apache-2.0 in its model card but ships no licence file; LICENSE is the canonical text from apache.org.
• Changes from the source: converted from PyTorch to Core AI (.aimodel) by Aether forge (recipe smollm3-3b@2). Weights are int8-linear-perblock32 (8-bit weights). The tokenizer files are the source's own.

Variants

Variant	Platform	Arch	Compute	Compiled	Download
macos-any-gpu	macos	any	gpu	no (specialized on first load)	3.28 GB
ios-h18p-gpu	ios	h18p	gpu	yes	3.29 GB

Verification

Every row is a record in verification/ about exactly these bytes (matched by bundle digest). Reference rows are strict T2 passes of the unquantized export on the same fixture, in verification/reference/.

Variant	Tier	Result	Detail	Device	OS build	Compute	Record
ios-h18p-gpu	T0	pass		iPhone18,2	24A437	target	552adde1
ios-h18p-gpu	T2	pass	19/19 strict; profile quantized-8bit; fixture edbd242c1ba166f2	iPhone18,2	24A437	target	68299013
macos-any-gpu	T0	pass		Mac17,6	26A428	target	485c762a
macos-any-gpu	T1	pass		Mac17,6	26A428	target	f9303bc7
macos-any-gpu	T2	pass	19/19 strict; 19/19 strict; profile quantized-8bit; fixture edbd242c1ba166f2	Mac17,6	26A428	target	275a91a6
unquantized reference (not published)	T2	pass	19/19 strict; 19/19 strict; profile strict; fixture edbd242c1ba166f2	Mac17,6	26A428	target	72c5e4b0

The quantized profile also requires: T2 strict on the unquantized reference export (met by the reference row).

Use

aether run smollm3-3b --prompt "Hello"

import Aether

let aether = try Aether()
let chat = try await aether.chat("smollm3-3b")
let reply = try await chat.respond(to: "Hello")
print(reply.text)