GPT-4's level of ability got about 300× cheaper in under two years, then stopped falling
The cheapest model matching GPT-4 on PhD-level science fell from $37.50 to $0.12 per million tokens by late 2024. Top models stay pricey.
Price of GPT-4-level performance
Cheapest non-reasoning model scoring at least GPT-4's level (≥33% on GPQA Diamond), USD per million tokens (3:1 input:output blend). Log scale.
What this measures: What it costs to use an AI at least as capable as the original GPT-4 (2023). A million tokens is roughly 750,000 words of input and output.
2023–2024 points are Epoch AI's. Points from late 2024 on are AI Tracker's extension using archived OpenRouter price lists and published GPQA Diamond scores (one provider route, not always the cheapest). The price of this capability level has flattened at roughly $0.05–0.13 since late 2024; the steep declines are now at higher capability levels (see the ARC-AGI cost chart). Per-token price is not cost per task.
Show data table and sources
| Date | Label | Value |
|---|---|---|
| 14 Mar 2023 | GPT-4 | $37.50 |
| 6 Nov 2023 | GPT-4 Turbo | $15 |
| 26 Feb 2024 | Mistral Large | $6 |
| 4 Mar 2024 | Claude 3 Haiku | $0.50 |
| 4 Nov 2024 | Gemini 1.5 Flash | $0.13 |
| 13 Dec 2024 | Phi 4 | $0.12 |
| 3 May 2025 | Gemma 3 12B | $0.063 |
| 1 Nov 2025 | Gemma 3 12B | $0.048 |
| 4 May 2026 | Mistral Small 3 | $0.058 |
| 1 Nov 2026 | Mistral Small 3 | $0.058 |
- LLM inference prices have fallen rapidly but unequally across tasks (Epoch AI)
- Lowest inference prices at fixed performance (CSV) (Epoch AI)
- OpenRouter models API (archived 2025-11-01) (OpenRouter via Internet Archive)
- OpenRouter models API (archived 2026-05-04) (OpenRouter via Internet Archive)
- Gemma 3 Technical Report (GPQA Diamond scores) (Google DeepMind)
Epoch AI tracked the cheapest model that could still reach a given score. The level it is easiest to follow is GPT-4’s score on PhD-level science questions. Matching it cost $37.50 per million tokens when GPT-4 launched in March 2023. By December 2024 a much smaller model, Phi 4, matched it for $0.12, a fall of about 40× a year. Across all the levels Epoch measured, prices fell between 9× and 900× a year.
Then it stopped. Extending Epoch’s series with archived price lists shows the cheapest route to this level has sat at about $0.05 to $0.13 per million tokens since late 2024, where it still is today.
The steep falls have moved up to harder levels of ability instead, which the cost of thought and ARC-AGI entries on this site track. A separate analysis by the investment firm a16z put the fall for a constant level of ability at about 10× a year.
What it means
Anything GPT-4 could do is now cheap enough to build into everyday software without much thought about the bill, which is why so many products added AI features after 2024. The caution is that this is one fixed level of ability getting cheaper, not AI in general: today’s best models cost far more, and the price of this particular level has not fallen for about two years. Cheap AI and the best AI are two different purchases.
Caveats
- Prices here mix input and output tokens 3:1, and leave out reasoning models, which use far more tokens per answer, so they are not what a user of a thinking model would pay.
- The figures after 2024 are AI Tracker’s own analysis, using one provider’s listed price; borderline cheaper models were left out.
- Test scores can be inflated when a model was trained on similar questions, so “matching GPT-4” may flatter the cheaper models.