M
MICROSOFT / microsoft/Aion-1.0-Instruct-2.3B

Aion 1.0 Instruct 2.3B

Microsoft Windows Applied Sciences Group text model published on November 18. It has about 2.29 billion parameters and MIT-licensed BF16 weights. Designed for summarization, rewriting, chat and document understanding in eight languages, including Inferama’s six languages.

01 / SUMMARY

The essential

ORIENTATIONSmall model for on-device text

Primary use stated or inferred cautiously from official documentation.

CONTEXT32,768 tokens

Maximum input capacity when published by the source.

EXITText; separate output maximum unpublished

Documented maximum generation limit.

ACCESSHugging Face · downloadable checkpoint

Verified availability channels.

MODEL TYPEText and reasoning

Primary functional family and verified specialties.

LICENSEMIT

Declared terms for API access, model weights, or self-hosting.

02 / TECHNICAL SHEET

Limits and integration

API IDmicrosoft/Aion-1.0-Instruct-2.3B
Model typeText and reasoning
Access modelOpen weights
LicenseMIT
DeploymentSelf-hosted
Release18 NOV 2026
Knowledge cutoffDecember 2025
EntranceText
ExitText
Context window32,768 tokens
maximum outputText; separate output maximum unpublished
ReasoningNot published
Published tools
Structured OutingsNot published
Batch processingNot published
Prompt cacheNot published
Fine-tuningNot published
Verified platformsHugging Face · Transformers · ONNX Runtime · WebGPU
BEST SUITED FOR

Local text-processing prototypes on compatible computers and browsers. The release adds an ONNX/WebGPU format with INT4 weights and FP16 activations and cache, around 1.3 GB. It is a deployment of the same model, not a duplicate profile.

WORTH MONITORING

The public version is not an already shipped Windows feature. It has the knowledge, reasoning and safety limitations of a small model. Transformers needs custom code: review it before loading. Local privacy also depends on the application, its telemetry and its connections.

03 / CAPABILITIES

What it can do

01

Orientation

Local text-processing prototypes on compatible computers and browsers. The release adds an ONNX/WebGPU format with INT4 weights and FP16 activations and cache, around 1.3 GB. It is a deployment of the same model, not a duplicate profile.

02

Context

32,768 tokens · Text; separate output maximum unpublished

03

Tools and integration

Not published

04

Access

Hugging Face · downloadable checkpoint

Input modalities
Text
04 / PRICES

Documented cost

ConceptWorthUnit/condition
Standard inputNo published official API rateStandard API rate
Cached inputNot publishedReading reused prefixes
Cache write or storageNot publishedThe condition varies by provider
Standard outputSelf-hosted infrastructure costMay include reasoning tokens
Batch inputNot publishedAsynchronous processing
Batch outputNot publishedAsynchronous processing

Weights and the ONNX format are published under MIT. No official API price is verified. Local execution consumes memory, energy and compute; file size is not total required memory. Quantization may change accuracy and performance.

i Prices change and may depend on level, region or context length. Check the source before making a decision.

05 / EVALUATIONS

How to read the results

“

Microsoft reports 55.15% MMLU and 83.90% IFEval. It uses lm-evaluation-harness when an implementation exists and the benchmark’s official implementation otherwise. IFEval averages prompt-level and instruction-level strict scores. These figures do not automatically apply to INT4 or every language.

BenchmarkResultMetricSource
MMLU55.15%Published scoreView source ↗
IFEval83.90%Mean of prompt-level and instruction-level strict accuracyView source ↗

Inferama only highlights a “best result” when the metric, test set, configuration, and date allow for an equivalent comparison. The supplier's figures are presented as claims from its own source.

06 / VERSIONS

Chronology

microsoft/Aion-1.0-Instruct-2.3B

BF16 checkpoint with custom code and MIT license.

microsoft/Aion-1.0-Instruct-2.3B-ONNX-WebGPU

ONNX INT4 format, revision 1677687f; does not add a model to the count.

07 / SOURCES

Traceability