Biggest Price Collapse

Models ranked by percentage price reduction from the GPT-3 launch baseline of $60 per million input tokens. This proxy uses current pricing vs. the original GPT-3 cost.

Methodology: (60 - current_input_price) / 60 * 100. Only active models included. GPT-3 launched at $60/1M input tokens in June 2020.

# Model Metric
1 Gemini 1.5 Flash ↓99.9% from $60
2 Llama 3.3 70B ↓99.8% from $60
3 GPT-4o mini ↓99.8% from $60
4 Gemini 2.0 Flash ↓99.8% from $60
5 Claude 3 Haiku ↓99.6% from $60
6 Mistral 7B ↓99.6% from $60
7 DeepSeek V3 ↓99.5% from $60
8 Qwen 2.5 72B ↓99.3% from $60
9 DeepSeek R1 ↓99.1% from $60
10 Mixtral 8x7B ↓98.8% from $60
11 Claude 3.5 Haiku ↓98.7% from $60
12 Llama 3 70B ↓98.5% from $60
13 Gemini 1.5 Pro ↓97.9% from $60
14 Grok 2 ↓96.7% from $60
15 GPT-4o ↓95.8% from $60
16 Claude 3.5 Sonnet ↓95.0% from $60
17 Claude Sonnet 4.6 ↓95.0% from $60
18 Mistral Large 2 ↓95.0% from $60
19 Command R+ ↓95.0% from $60
20 Llama 3.1 405B ↓91.7% from $60
21 o1 ↓75.0% from $60
22 Claude 3 Opus ↓75.0% from $60
← All leaderboards