HaiOK 中文站

AI Large Language Model Rankings

Data-driven ranking of frontier and open LLMs, with live per-token pricing from our automated daily pipeline (LiteLLM + OpenRouter sources).

As of 2026-09-07 · Source: HaiOK (haiokai.com) automated pipeline

TL;DR: For pay-as-you-go APIs, DeepSeek-class models are the cheapest; for frontier quality, GPT / Claude / Gemini lead; open-weight models offer the best price-performance. Prices below are per 1M tokens (USD).

Frontier & Open Models

#ModelVendorContextInputOutput
1 GPT-5 OpenAI 272K $1.25 $10.00
2 GPT-5 mini OpenAI 272K $0.25 $2.00
3 GPT-4.1 OpenAI 1.048M $2.00 $8.00
4 o3 OpenAI 200K $2.00 $8.00
5 Claude Opus 4.8 Anthropic 1M $5.00 $25.00
6 Claude Sonnet 4.6 Anthropic 1M $3.00 $15.00
7 Claude Haiku 4.5 Anthropic 200K $1.00 $5.00
8 Gemini 2.5 Pro Google 1.049M $1.25 $10.00
9 Gemini 2.5 Flash Google 1.049M $0.3 $2.50
10 Grok 4 xAI 131K $3.00 $15.00
11 Mistral Large (24.11) Mistral 128K $2.00 $6.00
12 DeepSeek V3 (chat) DeepSeek 131K $0.28 $0.42
13 DeepSeek R (reasoner) DeepSeek 131K $0.28 $0.42
14 GLM-4.5 Z.ai 128K $0.6 $2.20
15 Kimi K2.6 Moonshot AI 262K $0.95 $4.00
16 Llama 4 Maverick Meta 131K $0.2 $0.6

FAQ

Where does this data come from?

From our automated pipeline pulling public sources daily (LiteLLM, OpenRouter, GitHub API). As of 2026-09-07.

Is this an official ranking?

No. Rankings are third-party references, not official. Prices are reference prices — always check official pages.

Reference price · Rankings are community references, not official rankings. Prices are reference prices with update timestamps. Some outbound links may be affiliate links.