DeepInfra vs OpenRouter.
Prices checked
DeepInfra is cheaper on 32 of the 107 models both sell, the widest gap Llama 3.2 3B Instruct at 2¢ against $0.12 per M tokens, checked .
Where to start
DeepInfraall 126 models Serves open-weight models on its own GPUs, often at fp8 or fp4, and publishes the precision per model.
Cheapest of the shared models
2¢/M tokens
Llama 3.2 3B Instruct
OpenRouterall 297 models The reference router: charges each host's own price and routes open-weight models to the cheapest healthy host, so its price for an open model is often below the maker's.
Cheapest of the shared models
2.9¢/M tokens
Mistral Nemo
Side by side
| Fact | DeepInfra | OpenRouter |
|---|---|---|
| Kind of provider | Host | Router |
| Models we track | 126 | 297 |
| Cheaper of the two | 32 of 107 | 21 of 107 |
| Trains on your prompts | does not say | does not say |
| Keeps your requests | does not say | does not say |
| How you pay | does not say | does not say |
| Smallest top-up | does not say | does not say |
| Free tier | none listed | none listed |
| Where it runs | does not say | does not say |
Every model both sell
107 models, most widely sold first. Per M tokens, blended 3 in : 1 out.
| Model | DeepInfra | OpenRouter |
|---|---|---|
| GLM-5.3Z.ai · 1M context | $1.68 $0.90 in · $4 out | $0.40 7¢ in · $1.40 out |
| GLM-5.3-FlashZ.ai · 1M context | $0.24 $0.15 in · $0.50 out | $0.24 $0.15 in · $0.50 out |
| DeepSeek V4.1 Flash 1017DeepSeek · 1M context | $0.30 $0.20 in · $0.60 out | $0.26 $0.15 in · $0.60 out |
| DeepSeek V4 Flash 0731DeepSeek · 1M context | 9¢ 6¢ in · $0.18 out | 6.8¢ 4.3¢ in · $0.14 out |
| GLM-5.2Z.ai · 1M context | $1.16 $0.75 in · $2.40 out | $1.78 3.4¢ in · $7 out |
| Kimi K3Moonshot AI · 1M context | $5.70 $2.85 in · $14.25 out | $2.88 $0.24 in · $10.80 out |
| DeepSeek V4 ProDeepSeek · 1M context | $1.63 $1.30 in · $2.60 out | $0.99 $0.66 in · $1.98 out |
| Kimi K2.6Moonshot AI · 262k context | $1.44 $0.75 in · $3.50 out | $1.34 $0.65 in · $3.41 out |
| Qwen3.8 27BAlibaba Cloud · 1M context | $0.58 $0.15 in · $1.88 out | $0.34 3.9¢ in · $1.25 out |
| gpt-oss-120bOpenAI · 131k context | 7¢ 3.7¢ in · $0.17 out | $0.26 $0.15 in · $0.60 out |
| MiniMax-M3MiniMax · 1M context | $0.49 $0.28 in · $1.10 out | $0.52 $0.30 in · $1.20 out |
| Kimi K2.7 CodeMoonshot AI · 262k context | $1.36 $0.68 in · $3.40 out | $1.34 $0.67 in · $3.35 out |
| GLM-5.1Z.ai · 200k context | $1.66 $1.05 in · $3.50 out | $2.15 $1.40 in · $4.40 out |
| Claude Sonnet 4.6Anthropic · 1M context | $6 $3 in · $15 out | $6 $3 in · $15 out |
| Claude Haiku 4.5 (latest)Anthropic · 200k context | $2 $1 in · $5 out | $2 $1 in · $5 out |
| Claude Opus 4.8Anthropic · 1M context | $10 $5 in · $25 out | $10 $5 in · $25 out |
| Claude Opus 4.7Anthropic · 1M context | $10 $5 in · $25 out | $10 $5 in · $25 out |
| Claude Opus 5Anthropic · 1M context | $10 $5 in · $25 out | $10 $5 in · $25 out |
| Claude Fable 5Anthropic · 1M context | $20 $10 in · $50 out | $20 $10 in · $50 out |
| Claude Sonnet 5Anthropic · 1M context | $6 $3 in · $15 out | $4 $2 in · $10 out |
| Llama 3.3 70B InstructMeta · 131k context | $0.16 $0.10 in · $0.32 out | $0.16 $0.10 in · $0.32 out |
| DeepSeek V3.2DeepSeek · 164k context | $0.29 $0.26 in · $0.38 out | $0.29 $0.26 in · $0.38 out |
| MiniMax-M2.5MiniMax · 205k context | $0.40 $0.15 in · $1.15 out | $0.52 $0.30 in · $1.20 out |
| MiniMax-M2.7MiniMax · 205k context | $0.44 $0.25 in · $1 out | $0.52 $0.30 in · $1.20 out |
| gpt-oss-20bOpenAI · 131k context | 5.8¢ 3¢ in · $0.14 out | 3.6¢ 1.8¢ in · 9¢ out |
| Claude Opus 5.5Anthropic · 1M context | $8 $4 in · $20 out | $8 $4 in · $20 out |
| Kimi K2.5Moonshot AI · 262k context | $0.90 $0.45 in · $2.25 out | $1 $0.50 in · $2.50 out |
| Gemini 3.1 Pro PreviewGoogle · 1M context | $4.50 $2 in · $12 out | $4.50 $2 in · $12 out |
| Gemma 4 31B ITGoogle · 262k context | $0.15 9¢ in · $0.34 out | $0.25 $0.20 in · $0.40 out |
| Claude Sonnet 5.5Anthropic · 1M context | $4 $2 in · $10 out | $4 $2 in · $10 out |
| Qwen3.5 397B-A17BAlibaba Cloud · 262k context | $1.09 $0.45 in · $3 out | $1.27 $0.50 in · $3.60 out |
| Gemini 3.5 FlashGoogle · 1M context | $3.38 $1.50 in · $9 out | $3.38 $1.50 in · $9 out |
| Qwen3.6 35B-A3BAlibaba Cloud · 262k context | $0.31 $0.10 in · $0.95 out | $0.30 $0.10 in · $0.90 out |
| Gemini 3.7 FlashGoogle · 1M context | $1.50 $0.75 in · $3.75 out | $1.50 $0.75 in · $3.75 out |
| GLM-4.7Z.ai · 205k context | $0.74 $0.40 in · $1.75 out | $0.90 $0.54 in · $1.98 out |
| Gemini 2.5 FlashGoogle · 1M context | $0.85 $0.30 in · $2.50 out | $0.85 $0.30 in · $2.50 out |
| Gemini 3.1 Flash LiteGoogle · 1M context | $0.56 $0.25 in · $1.50 out | $0.56 $0.25 in · $1.50 out |
| Gemini 2.5 ProGoogle · 1M context | $3.44 $1.25 in · $10 out | $3.44 $1.25 in · $10 out |
| Gemma 4 26B A4B ITGoogle · 262k context | $0.14 7¢ in · $0.34 out | $0.11 6.8¢ in · $0.23 out |
| GLM-4.6Z.ai · 205k context | $0.88 $0.50 in · $2 out | $0.88 $0.50 in · $2 out |
The 40 most widely sold of 107 · every row in the open dataset
How we count
DeepInfra is cheaper on 32 of the 107 models both sell, OpenRouter on 21, and 54 cost the same at both. DeepInfra is a host and OpenRouter a router, so the two catalogs overlap without matching.
Each price is the cheapest that provider charges for the model, per M tokens, blended 3 in : 1 out. The smaller line under it is the price in and the price out.
A free tier is a limit rather than a price, and a price below the maker's own with nothing published to account for it is left out, so neither can be marked as the cheaper of the two.
Frequently asked
- Is DeepInfra cheaper than OpenRouter?
- On the 107 models both sell, DeepInfra is cheaper on 32, OpenRouter on 21, and 54 cost the same at both. The widest gap is Llama 3.2 3B Instruct: 2¢ per million tokens against $0.12. Prices are per M tokens, blended 3 in : 1 out, and read on 31 Oct 2026.
- Which models do DeepInfra and OpenRouter both sell?
- 107 of the models we track, led by GLM-5.3, GLM-5.3-Flash, DeepSeek V4.1 Flash 1017. DeepInfra sells 126 in all and OpenRouter 297, so most of each catalog has no counterpart at the other.
- Does DeepInfra or OpenRouter train on what you send it?
- DeepInfra does not say whether it trains on what you send it, and does not say how long it keeps requests. OpenRouter does not say whether it trains on what you send it, and does not say how long it keeps requests. A provider that may train on prompts is often how a price below the maker's own is paid for.
- How current are these prices?
- Both catalogs are read every 3 days. DeepInfra was last read 31 Oct 2026 and OpenRouter 31 Oct 2026; the whole index was refreshed 31 Oct 2026. Prices exclude tax, and a price below the maker's own with nothing published to account for it is left out of this table entirely.
Provider pages