Sulat.com
AI models
$10 off the fastest DeepSeek V4.1 Flash, Kimi K3 and GLM 5.3 from Synthetic
Kilo Gateway logo

Model details

AionLabs: Aion 3.5

AionLabs: Aion 3.5 is a text-to-text system designed around roleplaying and storytelling, built on the GLM family of foundation models from AionLabs. Rather than relying on a single generative pass, the model coordinates multiple specialized components in a collaborative generation process, letting each contribute to different aspects of a scene or narrative. It was released on September 23, 2026, and shares the same 262K-token context window as its sibling variant, Aion 3.5 Mini, fitting long-form interactive fiction and multi-turn roleplay scenarios where continuity matters.

In third-party benchmarking, Aion 3.5 stands out for exceptional reliability, posting a 100% success rate across eight evaluated tasks with no technical failures. It earns perfect accuracy on Hallucinations, General Knowledge, Reasoning, and Ethics benchmarks, and follows that up with 97.0% on Coding and 99.0% on Email Classification, while its Mathematics score lands at 92.9%. The primary weakness surfaces in Instruction Following, where it reaches only 65.0%, suggesting that complex, layered directives remain harder than open-ended reasoning. Overall the model is positioned for demanding creative and analytical applications where ethical grounding, factual steadiness, and consistent execution outweigh raw instruction adherence.

Kilo Gatewayaion-labs/aion-3.5

Quick Info

Powered by
Provider
Kilo Gateway
Model key
aion-labs/aion-3.5
Release date
Sep 23, 2026
Last updated
Sep 23, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost
$3.00
Output token cost
$6.00

Limits

Output tokens
32,768 tokens
Context window
262,144 tokens

Latest news about AionLabs: Aion 3.5

Kilo Gateway

CoverageBenchmark

A benchmark registry page documents AionLabs: Aion 3.5 as a multi-model roleplaying and storytelling system built on the GLM family, released September 23, 2026, with a 262K-token context window. It is listed at premium pricing of $3 per 1M input tokens, $6 per 1M output tokens, and $0.75 per 1M input cache-read tokens The page also enumerates supported parameters (Top P, Max Tokens, Temperature, Tool Choice, Response Format, Include Reasoning) and features (Tools, Response Format, Reasoning), and reports benchmark accuracy figures including 100% on Hallucinations, General Knowledge, Reasoning, and Ethics, 97.0% on Coding, and 99.0%

Videos about AionLabs: Aion 3.5