Aion Labs
Listed at https://aionlabs.ai
Overall Rank #44 ⭐ Consider
⚠️ Poland-based service with no mainland-China-optimized endpoint; China latency needs measurement. No regional data residency guarantees. | 🌍 International
💰 Token Pricing
| Type | Price | Note |
|---|---|---|
| Input | 按每百万 token 计费（2026-11-03 官网与 OpenRouter 一致）：aion-3.5-mini $0.70、aion-3.0-mini $0.70、aion-2.0 $0.80、aion-rp-llama-3.1-8b $0.80、aion-3.5 $3.00、aion-3.0 $3.00 | per million tokens |
| Output | 按每百万 token 计费（2026-11-03 官网与 OpenRouter 一致）：aion-3.5-mini $1.40、aion-3.0-mini $1.40、aion-2.0 $1.60、aion-rp-llama-3.1-8b $1.60、aion-3.5 $6.00、aion-3.0 $6.00 | per million tokens |
💡 Free Credits: Daily credit allowance free tier, no credit card required; rate-limited to 15 requests/min, 20K tokens/min, 20K tokens/day
🤖 Supported Models (6)
aion-3.5 (multi-model GLM system, 256K ctx, $3.00/$6.00, cached $0.75)aion-3.5-mini (multi-model GLM system, 256K ctx, $0.70/$1.40, cached $0.18)aion-3.0 (multi-model GLM/DeepSeek system, 128K ctx, $3.00/$6.00, cached $0.75)aion-3.0-mini (multi-model DeepSeek system, 128K ctx, $0.70/$1.40, cached $0.18)aion-2.0 (DeepSeek V3.2 variant for roleplay, 128K ctx, $0.80/$1.60, cached $0.20)aion-rp-llama-3.1-8b (Llama 3.1 8B fine-tune for RP, 32K ctx, $0.80/$1.60)
✨ Pros
- ✓Multi-model collaborative narrative system: multiple specialized models contribute to each response, optimized for roleplay and storytelling
- ✓Flagship Aion 3.5 (released 2026-09-23) offers 256K context window and adjustable reasoning effort (low/high/max)
- ✓75% cache discount on cached input ($0.18/M on Mini), ideal for roleplay apps that re-read world-building documents
- ✓All six models listed on OpenRouter with prices identical to the official docs
- ✓Email Agents: each agent has a dedicated email address and isolated compute environment, works over plain SMTP
- ✓Polish company (Deep Forge sp. z o.o.), Terms of Service available in English and Polish
⚠️ Cons
- ×No tool calling or function calling on any model; unsuitable for agentic workflows
- ×Flagship tier at $3.00/$6.00 is expensive for pure narrative scenarios; general-purpose models available at similar price
- ×Free tier is only 15 requests/min and 20K tokens/min, enough for evaluation only
- ×No Batch API, no fine-tuning endpoint, no committed-spend discount
- ×Email Agents require contacting sales; no self-serve evaluation
- ×No SOC 2 compliance documentation or regional data residency guarantees
🎯 Best For
Teams building interactive fiction, roleplay applications, and narrative games; multi-session story arcs needing 256K long-context narrative coherence; enterprises that live in email and want to deploy AI workers via Email Agents