Which LLM is best — and cheapest?
Traceable side-by-side comparison of performance, pricing and subscription plans — including Chinese frontier models (DeepSeek, Doubao, Qwen, GLM, Kimi). Cost calculator, plan comparison, every figure sourced and dated.
Find your model
1. What will you use it for?
Not sure where to start? Pick by use case
Coding & reasoning
GPT-5.5 Pro
Highest-tier GPT-5 with extended reasoning; pro mode for hardest tasks.
Daily chat (free)
Gemini 3 Flash-Lite
Google's most economical model; 1M context at the lowest price.
Long documents
Llama 4 Scout
Compact open-weight Llama 4 with extreme 10M token context window.
Cheapest
DeepSeek V4 Flash
Cheapest high-volume chat and light tasks
Vision / multimodal
Gemini 3 Pro
Very long documents (1M+), native multimodal, strong reasoning
Agent & tool use
Claude Opus 4.8
Long-horizon agents, complex multi-step tasks, careful writing
Top models
Ranked by value-for-money (intelligence ÷ price)
Why does a cheaper model rank first? Value = intelligence ÷ price (log-adjusted) — prevents ultra-cheap models from dominating. DeepSeek is both cheap and strong. Sort by intelligence to see the strongest.
| # | Model | Intelligence | Input price |
|---|---|---|---|
| 1 | DeepSeek V4 Flash | 4.2/10Above average | $0.070 |
| 2 | Doubao Seed 1.6● Free tier | 4.5/10Above average | $0.11 |
| 3 | Doubao 2.0● Free tier | 4.9/10Above average | $0.15 |
| 4 | Llama 4 Scout | 2.8/10Entry-level | $0.10 |
| 5 | Llama 4 Maverick | 3.0/10Entry-level | $0.15 |
| 6 | Grok 4.1 Fast | 4.3/10Above average | $0.20 |
| 7 | DeepSeek V4 Pro● Free tier | 5.2/10High | $0.42 |
| 8 | Kimi K2.5● Free tier | 5.0/10High | $0.38 |
Artificial Analysis — Intelligence Index· 2026-06-17= Intelligence ÷ price (log-adjusted to prevent ultra-cheap domination) · higher = better value. Max across all models ≈ 195.2.
Models with a free tier
Try before you buy