The cheapest LLM API for your workload
A model that is cheap on input can be dear on output, so the cheapest API depends on how much you send and how much comes back. Set your workload and compare the monthly bill of 35 models.
≈ 60M input / 17.5M output tokens a month
What it costs a month, by model
gpt-oss-120b is cheapest for this workload at $1.38 a month; GPT-5 Pro costs about 2,178 times more.
- 1gpt-oss-120bCheapest$1.38
- 2Mistral Nemo$2.27
- 3gpt-oss-20b$4.07
- 4Llama 3.1 8B Instruct$4.40
- 5Nova Micro 1.0$4.55
- 6Command R7B (12-2024)$4.90
- 7Gemma 3 12B$5.63
- 8Nova Lite 1.0$7.80
- 13Qwen3 32B$9.70
- 17Gemini 2.5 Flash Lite$13.00
- 22Gemini 2.5 Flash$61.75
- 26Grok 4.7$225
- 31Claude Sonnet 4.5$443
- 35GPT-5 Pro$3,000
All models (35)
| # | Model | In / 1M | Out / 1M | Your cost / mo |
|---|---|---|---|---|
| 1 | gpt-oss-120b | $0.011 | $0.041 | $1.38 |
| 2 | Mistral Nemo | $0.029 | $0.03 | $2.27 |
| 3 | gpt-oss-20b | $0.03 | $0.13 | $4.07 |
| 4 | Llama 3.1 8B Instruct | $0.05 | $0.08 | $4.40 |
| 5 | Nova Micro 1.0 | $0.035 | $0.14 | $4.55 |
| 6 | Command R7B (12-2024) | $0.038 | $0.15 | $4.90 |
| 7 | Gemma 3 12B | $0.05 | $0.15 | $5.63 |
| 8 | Nova Lite 1.0 | $0.06 | $0.24 | $7.80 |
| 9 | Mistral Small 3.2 24B | $0.075 | $0.20 | $8.00 |
| 10 | Qwen3.5-Flash | $0.065 | $0.26 | $8.45 |
| 11 | DeepSeek V4 Flash 0423 | $0.091 | $0.182 | $8.65 |
| 12 | Qwen2.5 7B Instruct | $0.10 | $0.20 | $9.50 |
| 13 | Qwen3 32B | $0.08 | $0.28 | $9.70 |
| 14 | GPT-5 Nano | $0.05 | $0.40 | $10.00 |
| 15 | Gemma 3 27B | $0.08 | $0.35 | $10.93 |
| 16 | GPT-4.1 Nano | $0.10 | $0.40 | $13.00 |
| 17 | Gemini 2.5 Flash Lite | $0.10 | $0.40 | $13.00 |
| 18 | Claude Haiku 5.5 | $0.10 | $0.50 | $14.75 |
| 19 | GPT-5 Mini | $0.25 | $2.00 | $50.00 |
| 20 | GPT-4.1 Mini | $0.40 | $1.60 | $52.00 |
| 21 | Mistral Medium 3 | $0.40 | $2.00 | $59.00 |
| 22 | Gemini 2.5 Flash | $0.30 | $2.50 | $61.75 |
| 23 | R1 0528 | $0.50 | $2.15 | $67.63 |
| 24 | Kimi K2 0711 | $0.57 | $2.30 | $74.45 |
| 25 | Grok 4.3 | $1.25 | $2.50 | $119 |
| 26 | Grok 4.7 | $2.00 | $6.00 | $225 |
| 27 | GPT-5 | $1.25 | $10.00 | $250 |
| 28 | GPT-5.1 | $1.25 | $10.00 | $250 |
| 29 | Gemini 2.5 Pro | $1.25 | $10.00 | $250 |
| 30 | Claude Sonnet 5.5 | $2.00 | $10.00 | $295 |
| 31 | Claude Sonnet 4.5 | $3.00 | $15.00 | $443 |
| 32 | Claude Opus 5.5 | $4.00 | $20.00 | $590 |
| 33 | Claude Opus 5 | $5.00 | $25.00 | $738 |
| 34 | GPT-5.5 | $5.00 | $30.00 | $825 |
| 35 | GPT-5 Pro | $15.00 | $120 | $3,000 |
Input and output prices are per one million tokens. "Your cost / mo" is your workload priced at those two rates.
How to read this
The order changes with the job. Classification sends long inputs and gets short answers, so the input price decides it; a chat assistant writes more, so the output price counts for more. Pick the workload closest to yours, then compare the few cheapest on quality before you commit. Price only tells you where to start looking.
Prices read on 2026-11-05: OpenRouter model list · all model prices · the full comparison
Paying more than this, on the wrong model?
We route each task to the smallest model that holds the quality and wire it into the work you already do. Book a free call and we'll price your workload properly.
Book a callBefore we count this visit
Analytics and ad measurement tell us which pages and which of our ads do their job: Google Analytics and Google Ads, the Meta pixel, Yandex Metrica, and one first-party id of our own. Nothing is stored until you answer, and declining changes nothing about the site.