Aion Labs Launches Aion 3.5 Mini, a $0.70/M-Token Roleplaying Model with 262K Context

Aion Labs has released Aion 3.5 Mini, a smaller and cheaper version of its Aion 3.5 roleplaying and storytelling model, according to a listing on OpenRouter. The model is priced at $0.70 per 1 million input tokens and $1.40 per 1 million output tokens, with a 262,000 token context window.

Aion 3.5 Mini is built on the GLM family of models and uses what Aion Labs calls a "collaborative generation process," in which multiple specialized models each contribute to a single response. According to Aion Labs, this multi-model architecture is designed to produce stronger narrative structure and more compelling tension and conflict in generated text compared to single-model systems.

The Mini variant is positioned as a lower-cost sibling to the full Aion 3.5 model, which shares the same 262K context window and architecture but is priced substantially higher at $3 per 1M input tokens and $6 per 1M output tokens. That makes Aion 3.5 Mini roughly 4x cheaper on input and output pricing while retaining the same context length.

On OpenRouter, the model shows a median latency of 0.77 seconds and throughput of 42 tokens per second, with 97.14% availability over the trailing 24-hour window at the time of listing. Cached input tokens are priced at $0.18 per 1 million.

This release extends a lineup that already includes Aion-3.0-Mini (DeepSeek-based, 131K context, same $0.70/$1.40 pricing), Aion-3.0 (GLM-based, 131K context, $3/$6), Aion-2.0 (DeepSeek V3.2 variant, 131K context, $0.80/$1.60), and earlier reasoning-focused models Aion-1.0 and Aion-1.0-Mini built on DeepSeek-R1. Aion Labs has also released Aion-RP 1.0, a fine-tuned Llama-3.1-8B model the company says ranks highest in the character evaluation portion of the RPBench-Auto benchmark, an Arena-Hard-Auto variant for roleplaying.

No independent benchmark scores for Aion 3.5 Mini have been published. Aion Labs has not disclosed training data cutoff dates or parameter counts for the model.

What this means

Aion Labs continues to carve out a niche in AI-driven roleplaying and interactive fiction rather than competing on general-purpose reasoning or coding benchmarks. The Mini/full-size pairing strategy — offering a cheaper model at the same context length as its flagship — mirrors a pattern common across the industry, letting cost-sensitive developers and hobbyist app builders access long-context storytelling without paying premium per-token rates. At $0.70/$1.40 per 1M tokens, Aion 3.5 Mini undercuts many general-purpose models with comparable context windows, though its narrow specialization in roleplaying and multi-model generation pipeline may add latency or cost variability compared to single-model competitors. The lack of independently verified benchmarks means buyers should treat Aion Labs' narrative-quality claims as unverified until third-party evaluation data emerges.