AI Economics
What it costs to run AI: language models per token, images and video per unit, and the GPUs underneath. 1,778 models from 133 hosts, read every hour.
What a strong model costs, over time
- 70% on GPQA Diamond
- At or above GPT-4 on LMArena
The numbers
Each figure is worked out from the prices and scores on this site. Open one to see the formula and the sources.
- A 70% GPQA Diamond modelShare$0.054
The cheapest model that scores 70% or more on GPQA Diamond costs $0.054 per 1M tokens, blended 3 input to 1 output, on 2026-10-23.
How this is worked out
Cheapest current blended price among models with an Epoch AI GPQA Diamond mean score of 0.70 or more.
Source: Epoch AI, OpenRouter, LiteLLM price map. As of 2026-10-23. Based on 125 data points.
- At or above GPT-4 on LMArenaShare$0.025
The cheapest model rated at or above GPT-4 (June 2023) on the LMArena text leaderboard costs $0.025 per 1M tokens, blended 3 input to 1 output, ×1500 less than in Dec 2023. Small models such as Llama 3.1 8B clear this bar.
How this is worked out
Cheapest blended price (3 input : 1 output tokens, per 1M) among models whose LMArena text rating is at least that of gpt-4-0613 on the same leaderboard.
Source: LiteLLM price map, OpenRouter, LMArena. As of 2026-10-23. Based on 233 data points.
- Same model, different hostShare×2.0
The same open model costs ×2.0 as much at the dearest host as at the cheapest, at the median across 170 models, comparing hosts that serve the same stated precision.
How this is worked out
For each model with three or more hosts serving the same stated precision (fp8 with fp8, and so on): dearest ÷ cheapest list price per 1M input tokens. Median across models. Promotional prices count as listed.
Source: OpenRouter. Based on 170 data points.
- US models cost less, same qualityShare×1.9 gap
At 60% on GPQA Diamond, the cheapest US model costs $0.028 per 1M tokens; the cheapest Chinese model $0.054.
How this is worked out
Cheapest blended price per developer country among models with an Epoch AI GPQA Diamond score of 0.60 or more (at least three models per country). Country is the developer’s, from Epoch AI.
Source: Epoch AI, OpenRouter, LiteLLM price map. As of 2026-10-23. Based on 143 data points.
- A top-10 imageShare$0.019
An image from one of the ten best-rated image models costs $0.019 at the median.
How this is worked out
Median of the cheapest per-image price (standard size) among the ten highest-rated LMArena text-to-image models.
Source: LiteLLM price map, LMArena. As of 2026-10-23. Based on 2 data points.
- A 10-second 1080p clipShare$0.80
The cheapest 10-second 1080p AI video costs $0.80 ($0.08 a second).
How this is worked out
Cheapest per-second price at 1080p across video models × 10 seconds.
Source: LiteLLM price map. As of 2026-10-23. Based on 25 data points.
- H100 indexShare$9,087/mo
Renting one H100 costs $9,087 a month at the median of every on-demand offer.
How this is worked out
Median on-demand price per H100-hour across every cloud and GPU marketplace offer × 730 hours.
Source: Cloudadapt GPU prices. As of 2026-10-23.
- Cached inputShare90% off
Input tokens served from the prompt cache cost 90% less than fresh ones, at the median.
How this is worked out
Cached-input price ÷ list input price, per model and host that publish both. One minus the median.
Source: OpenRouter, LiteLLM price map. Based on 2,420 data points.
- Off-peak hoursShare50% off
DeepSeek charges up to 50% less in its off-peak hours (Beijing time).
How this is worked out
One minus the off-peak list price over the peak list price, per DeepSeek model and token type, in CNY.
Source: DeepSeek pricing. Based on 6 data points.
- Batch jobsShare50% off
Batch requests cost 50% less than live ones, at the median, where a host offers them.
How this is worked out
Batch input price ÷ list input price per model and host. One minus the median.
Source: LiteLLM price map. Based on 422 data points.
- List-price movesShare210 down, 105 up
In Sept 2026, 210 AI list prices went down and 105 went up.
How this is worked out
Count of list-price changes in the month across every model, host and unit, by direction.
Source: LiteLLM price map, OpenRouter. As of 2026-09. Based on 315 data points.
- Open weights cost less, same qualityShare×2.4 gap
At 60% on GPQA Diamond, the cheapest open-weights model costs $0.028 per 1M tokens and the cheapest closed one $0.069.
How this is worked out
Cheapest blended price among models with an Epoch AI GPQA Diamond score of 0.60 or more, grouped by whether the weights are published.
Source: Epoch AI, OpenRouter, LiteLLM price map. As of 2026-10-23. Based on 123 data points.
Explore
Latest list-price changes
| Model | Host | Unit | Before | Now | Change | Date |
|---|---|---|---|---|---|---|
| GLM 5.3 | Morph | $0.322 | $0.306 | −5% | ||
| DeepSeek V4 Pro 0813 | DeepSeek | $1.98 | $3.96 | +100% | ||
| DeepSeek V4.1 Flash 1017 | Morph | $0.265 | $0.288 | +9% | ||
| GLM 5.2 | Morph | $0.322 | $0.306 | −5% | ||
| DeepSeek V4 Flash 0731 | Wafer | $0.12 | $0.14 | +17% | ||
| DeepSeek V4.1 Flash | Morph | $0.265 | $0.288 | +9% | ||
| DeepSeek V4.1 Flash 1017 | DeepSeek | $0.60 | $1.20 | +100% | ||
| DeepSeek V4.1 Flash 1017 | DeepSeek | $0.15 | $0.30 | +100% | ||
| DeepSeek V4 Pro 0813 | DeepSeek | $0.66 | $1.32 | +100% | ||
| DeepSeek V4 Flash 0731 | Relace | $0.16 | $0.157 | −2% |
Quality data: LMArena (CC BY 4.0), Epoch AI (CC BY 4.0). Prices: OpenRouter, LiteLLM (MIT), provider pages. List prices in US dollars before discounts and taxes; other currencies converted at the European Central Bank reference rate. The figures on these pages are free to reuse under CC BY 4.0 with a link to this page.