Best LLMs for Multi-step Planning
10 models · ranked by real Multi-step Planning usage
Key facts · week of 2026-08-10
The top 3 models (DeepSeek: DeepSeek V4 Flash 0423, DeepSeek: DeepSeek V4 Flash 0731, Tencent: Hy3) account for 26.7% of all usage.
§This week's biggest riser is DeepSeek: DeepSeek V4 Flash 0731, up from #6 to #2.
§The cheapest model in the usage top 10 is OpenAI: GPT-5.6 Luna (batch) at $0.09999999999999999 per 1M input tokens (usage rank #6).
§The largest context window in the top 10 is Xiaomi: MiMo-V2.5 at 1.1M tokens.
§
Cite as: AI Tier List, LLM Usage Rankings (2026-08-10), www.aitierlist.xyz/en/models/agent-multi-step-planning — data: OpenRouter
Which model should you pick?
- When quality comes first
- MoonshotAI: Kimi K3
- Artificial Analysis agentic benchmark score 54.3 — highest in this ranking
- The proven default
- DeepSeek: DeepSeek V4 Flash 0423
- 9.9% usage share — #1 in real usage
- When cost matters
- NVIDIA: Nemotron 3 Ultra (free)
- Cheapest in the top 15 — Free/1M input
- For long docs & codebases
- Xiaomi: MiMo-V2.5
- 1.1M token context window
- To bet on the trending choice
- DeepSeek: DeepSeek V4 Flash 0731
- Up 4 spots (#6 → #2)
Full ranking · 10 models
| Rank | Model | Share |
|---|---|---|
1 | DeepSeek: DeepSeek V4 Flash 0423 | |
2 | DeepSeek: DeepSeek V4 Flash 0731 | |
3 | Tencent: Hy3 | |
4 | Xiaomi: MiMo-V2.5 | |
5 | Z.ai: GLM 5.2 (batch) | |
6 | OpenAI: GPT-5.6 Luna (batch) | |
7 | DeepSeek: DeepSeek V4 Pro | |
8 | MiniMax: MiniMax M3 (batch) | |
9 | NVIDIA: Nemotron 3 Ultra (free) | |
10 | MoonshotAI: Kimi K3 |
Model Comparisons
- DeepSeek: DeepSeek V4 Flash 0423 vs DeepSeek: DeepSeek V4 Flash 0731
- DeepSeek: DeepSeek V4 Flash 0423 vs Tencent: Hy3
- DeepSeek: DeepSeek V4 Flash 0423 vs Xiaomi: MiMo-V2.5
- DeepSeek: DeepSeek V4 Flash 0731 vs Tencent: Hy3
- DeepSeek: DeepSeek V4 Flash 0731 vs Xiaomi: MiMo-V2.5
- Tencent: Hy3 vs Xiaomi: MiMo-V2.5
- Full LLM API pricing →
Related AI Tool Tier Lists
Get notified when Multi-step Planning rankings move
This ranking refreshes every week. We email you when it moves — once a week, no spam.
FAQ
What is the most-used LLM for Multi-step Planning?
DeepSeek: DeepSeek V4 Flash 0423 by deepseek ranks first for Multi-step Planning by real token usage. It accounts for roughly 9.9% of tokens in this segment. DeepSeek: DeepSeek V4 Flash 0731, Tencent: Hy3, Xiaomi: MiMo-V2.5 follow.
How are these rankings calculated?
Models are ranked by tokens actually consumed through OpenRouter, not by benchmark scores. This page counts only usage tagged as Multi-step Planning tasks. It refreshes weekly and currently tracks 10 models.
Which is the cheapest LLM for Multi-step Planning?
Among the paid models in this ranking, OpenAI: GPT-5.6 Luna (batch) is cheapest at $0.10 per 1M input tokens. 1 models in the list are free to use.
Is DeepSeek: DeepSeek V4 Flash 0423 still the most-used LLM for Multi-step Planning right now?
Yes. As of the week of 2026-08-10, DeepSeek: DeepSeek V4 Flash 0423 is #1 by real usage. It was #1 last week too. The ranking refreshes weekly, so this page changes as soon as it does.
Which model has the longest context window?
Xiaomi: MiMo-V2.5 has the longest context window in this ranking at 1.1M tokens.
Token usage and share data aggregated by OpenRouter, refreshed weekly. Prices are USD per 1M tokens. Benchmark scores come from Artificial Analysis; models it hasn't measured show —.