zamba2-1.2b-instruct

A Core AI bundle of Zyphra/Zamba2-1.2B-Instruct-v2 for the Aether SDK (iOS and macOS 27+).

• Source: Zyphra/Zamba2-1.2B-Instruct-v2 at revision 960222dd212c071e2b7e3734573ee5297b1c075d, licence Apache-2.0. The licence is included as LICENSE. Zyphra/Zamba2-1.2B-Instruct-v2 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 zamba2-1.2b-instruct@1). Weights are int8-linear-perblock32 (8-bit weights). The tokenizer files are the source's own.

Variants

Variant	Platform	Arch	Compute	Compiled	Assets	Download
macos-any-gpu	macos	any	gpu	no (specialized on first load)	zamba2_1_2b_instruct.aimodel 1.29 GB	1.3 GB
ios-any-gpu	ios	any	gpu	no (specialized on first load)	zamba2_1_2b_instruct.aimodel 1.29 GB	1.3 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-any-gpu	T0	pass		iPhone18,2	24A446	target	31aefeaf
ios-any-gpu	T2	pass	20/20 strict; profile quantized-8bit; fixture adbdef8a32ca3c1b	iPhone18,2	24A446	target	dd9e2cec
ios-any-gpu	T3	pass	copy-fidelity-v1; 80.0% vs reference 80.0%; 50 items	iPhone18,2	24A446	target	8aefdd45
macos-any-gpu	T0	pass		Mac17,6	26A434	target	f11695a8
macos-any-gpu	T1	pass		Mac17,6	26A434	target	cc6fad7c
macos-any-gpu	T2	pass	20/20 strict; 20/20 strict; profile quantized-8bit; fixture adbdef8a32ca3c1b	Mac17,6	26A434	target	693a2248
macos-any-gpu	T3	pass	gsm8k-test-500; 40.4% vs reference 40.4%; 500 items	Mac17,6	26A434	target	4722db3c
macos-any-gpu	T3	pass	copy-fidelity-v1; 80.0% vs reference 80.0%; 50 items	Mac17,6	26A434	target	b0bf808b
unquantized reference (not published)	T2	pass	20/20 strict; 20/20 strict; profile strict; fixture adbdef8a32ca3c1b	Mac17,6	26A434	target	b307cd38

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

Use

aether run zamba2-1.2b-instruct --prompt "Hello"

import Aether

let aether = try Aether()
let chat = try await aether.chat("zamba2-1.2b-instruct")
let reply = try await chat.respond(to: "Hello")
print(reply.text)