DeepInfra model pricing.
prices as of · 106 models · 180 prices
Serves open-weight models on its own GPUs, often at fp8 or fp4, and publishes the precision per model.
DeepInfra is a host selling 106 models, from $0.08 per million tokens for Llama 3.2 3B Instruct, read on 16 Sept 2026.
Kind of provider
Host
based in US
Where it runs
not stated
no regions published
How you pay
not stated
no payment terms published
Free tier
none published
nothing given away
Where DeepInfra's prices land, and why
DeepInfra publishes 180 prices we can compare, and 84 of them are below the list price for that model. 79 of those have a published reason behind them, so they rank like any other price. The other 3 sit below list on models whose weights are closed, with nothing published to account for the gap, so they are listed here in full and ranked nowhere on the site. Nobody here has read its terms yet, so treat what it says about itself as its own claim.
DeepInfra does not say whether it trains on what you send it. It does not say how long it keeps requests.
not yet checked: terms · privacy policy ↗ · status page ↗
Models to start with
all 180 prices ↓- DeepSeek V4 Flash 0731DeepSeek · 1M context · 97 providers sell it$0.09 / $0.1857% under listbuy at DeepInfra ↗
- DeepSeek V4 Flash 0731DeepSeek · 1M context · 97 providers sell it$0.09 / $0.1857% under listbuy at DeepInfra ↗
- GLM-5.2Z.ai · 1M context · 94 providers sell it$0.75 / $2.4046% under listbuy at DeepInfra ↗
- GLM-5.2Z.ai · 1M context · 94 providers sell it$0.75 / $2.4046% under listbuy at DeepInfra ↗
- DeepSeek V4 ProDeepSeek · 1M context · 84 providers sell it$1.30 / $2.60199% above listbuy at DeepInfra ↗
Everything DeepInfra sells, against the list price
Cheapest first inside each group. The share is this row against the price the lab that trained the model charges.
Cheaper, and we know why
79 pricesBelow list, with a published reason: hosting the open weights, a regional price list, credits, a stated discount.
- Llama 3.2 3B Instruct
Meta · 131k context
Hosts the open weights on its own servers, so the price is its own, not a resale. Precision not disclosed.
$0.02 / $0.02in / out per M
0.17×of list price
- Mistral Nemo
Mistral AI · 128k context
Serves the open weights at fp8, a lower precision than the maker's, so not quite the same product.
$0.02 / $0.03in / out per M
0.14×of list price
- Llama 3.1 8B Instruct
Meta · 131k context
Serves the open weights at fp8, a lower precision than the maker's, so not quite the same product.
$0.02 / $0.04in / out per M
0.43×of list price
- Llama 3.1 8B Instruct
Meta · 131k context
Hosts the open weights on its own servers, so the price is its own, not a resale. Precision not disclosed.
$0.02 / $0.05in / out per M
0.48×of list price
- Qwen2.5 7B Instruct
Alibaba Cloud · 131k context
Hosts the open weights on its own servers, so the price is its own, not a resale. Precision not disclosed.
$0.04 / $0.10in / out per M
0.18×of list price
- Mistral Small 3.2 24B
Mistral AI · 256k context
Hosts the open weights on its own servers, so the price is its own, not a resale. Precision not disclosed.
$0.07 / $0.20in / out per M
0.80×of list price
- Mistral Small 3.2 24B
Mistral AI · 256k context
Serves the open weights at fp8, a lower precision than the maker's, so not quite the same product.
$0.07 / $0.20in / out per M
0.80×of list price
- DeepSeek V4 Flash 0731
DeepSeek · 1M context
Hosts the open weights on its own servers, so the price is its own, not a resale. Precision not disclosed.
$0.09 / $0.18in / out per M
0.43×of list price
- DeepSeek V4 Flash 0731
DeepSeek · 1M context
Serves the open weights at fp8, a lower precision than the maker's, so not quite the same product.
$0.09 / $0.18in / out per M
0.43×of list price
- Qwen3 32B
Alibaba Cloud · 131k context
Hosts the open weights on its own servers, so the price is its own, not a resale. Precision not disclosed.
$0.08 / $0.28in / out per M
0.11×of list price
- Qwen3 32B
Alibaba Cloud · 131k context
Serves the open weights at fp8, a lower precision than the maker's, so not quite the same product.
$0.08 / $0.28in / out per M
0.11×of list price
- Gemma 4 26B A4B IT
Google · 262k context
Hosts the open weights on its own servers, so the price is its own, not a resale. Precision not disclosed.
$0.07 / $0.34in / out per M
0.52×of list price
- Gemma 4 26B A4B IT
Google · 262k context
Serves the open weights at fp8, a lower precision than the maker's, so not quite the same product.
$0.07 / $0.34in / out per M
0.52×of list price
- Qwen3 14B
Alibaba Cloud · 131k context
Hosts the open weights on its own servers, so the price is its own, not a resale. Precision not disclosed.
$0.12 / $0.24in / out per M
0.24×of list price
- Qwen3 14B
Alibaba Cloud · 131k context
Serves the open weights at fp8, a lower precision than the maker's, so not quite the same product.
$0.12 / $0.24in / out per M
0.24×of list price
- Nemotron 3 Super
NVIDIA · 262k context
Hosts the open weights on its own servers, so the price is its own, not a resale. Precision bf16.
$0.09 / $0.40in / out per M
0.47×of list price
- Qwen3 235B-A22B
Alibaba Cloud · 131k context
Hosts the open weights on its own servers, so the price is its own, not a resale. Precision not disclosed.
$0.09 / $0.55in / out per M
0.17×of list price
- Qwen3 235B-A22B
Alibaba Cloud · 131k context
Serves the open weights at fp8, a lower precision than the maker's, so not quite the same product.
$0.09 / $0.55in / out per M
0.17×of list price
- Qwen3 30B A3B Instruct 2507
Alibaba Cloud · 262k context
8% under the list price, an ordinary reseller margin.
$0.12 / $0.50in / out per M
0.95×of list price
- Qwen3 30B A3B Instruct 2507
Alibaba Cloud · 262k context
8% under the list price, an ordinary reseller margin.
$0.12 / $0.50in / out per M
0.95×of list price
- Qwen3-VL 30B-A3B
Alibaba Cloud · 131k context
Hosts the open weights on its own servers, so the price is its own, not a resale. Precision not disclosed.
$0.15 / $0.60in / out per M
0.75×of list price
- Qwen3-VL 30B-A3B
Alibaba Cloud · 131k context
Serves the open weights at fp8, a lower precision than the maker's, so not quite the same product.
$0.15 / $0.60in / out per M
0.75×of list price
$0.26 / $0.38in / out per M
0.94×of list price
$0.26 / $0.38in / out per M
0.94×of list price
- R1 Distill Llama 70B
DeepSeek · 8k context
Hosts the open weights on its own servers, so the price is its own, not a resale. Precision not disclosed.
$0.20 / $0.60in / out per M
0.38×of list price
- Qwen3.6 35B-A3B
Alibaba Cloud · 262k context
Hosts the open weights on its own servers, so the price is its own, not a resale. Precision not disclosed.
$0.10 / $0.95in / out per M
0.56×of list price
- Qwen3.6 35B-A3B
Alibaba Cloud · 262k context
Serves the open weights at fp8, a lower precision than the maker's, so not quite the same product.
$0.10 / $0.95in / out per M
0.56×of list price
- Qwen3-Next 80B-A3B Instruct
Alibaba Cloud · 131k context
Hosts the open weights on its own servers, so the price is its own, not a resale. Precision not disclosed.
$0.09 / $1.10in / out per M
0.39×of list price
- Qwen3-Next 80B-A3B Instruct
Alibaba Cloud · 131k context
Serves the open weights at fp8, a lower precision than the maker's, so not quite the same product.
$0.09 / $1.10in / out per M
0.39×of list price
- Qwen3.5 35B-A3B
Alibaba Cloud · 262k context
Hosts the open weights on its own servers, so the price is its own, not a resale. Precision not disclosed.
$0.14 / $1in / out per M
0.52×of list price
showing the 30 cheapest of 79 · every row in the open dataset
Same price as the maker
98 pricesThe same price the maker charges, sold by someone else.
$0.03 / $0.14in / out per M
1.05×of list price
$0.03 / $0.14in / out per M
1.05×of list price
$0.03 / $0.14in / out per M
1.05×of list price
$0.05 / $0.08in / out per M
1.00×of list price
$0.05 / $0.08in / out per M
1.00×of list price
$0.05 / $0.10in / out per M
1.00×of list price
$0.05 / $0.10in / out per M
1.00×of list price
$0.05 / $0.10in / out per M
1.00×of list price
$0.04 / $0.17in / out per M
1.00×of list price
$0.04 / $0.17in / out per M
1.00×of list price
$0.04 / $0.17in / out per M
1.00×of list price
$0.05 / $0.15in / out per M
1.00×of list price
$0.05 / $0.15in / out per M
1.00×of list price
$0.05 / $0.15in / out per M
1.00×of list price
$0.05 / $0.20in / out per M
1.00×of list price
$0.05 / $0.20in / out per M
1.00×of list price
$0.07 / $0.14in / out per M
1.00×of list price
$0.07 / $0.14in / out per M
1.00×of list price
$0.07 / $0.14in / out per M
1.00×of list price
$0.06 / $0.18in / out per M
2.86×of list price
$0.06 / $0.18in / out per M
2.86×of list price
$0.06 / $0.18in / out per M
1.00×of list price
$0.06 / $0.18in / out per M
1.00×of list price
$0.06 / $0.18in / out per M
1.00×of list price
$0.06 / $0.18in / out per M
1.00×of list price
$0.08 / $0.16in / out per M
0.58×of list price
$0.08 / $0.16in / out per M
0.58×of list price
$0.08 / $0.20in / out per M
1.00×of list price
$0.10 / $0.15in / out per M
1.00×of list price
$0.10 / $0.15in / out per M
1.00×of list price
showing the 30 cheapest of 98 · every row in the open dataset
Cheaper, and nobody says why
3 prices · never rankedBelow list on a closed model with nothing to account for it. Shown, never ranked, never a pick.
- Mistral Nemo
Mistral AI · 128k context
0.13× the list price, from one source and far below what hosting the weights costs. Listed until a second source agrees.
$0.02 / $0.03in / out per M
0.14×of list price
- Qwen3.8 Flash
Alibaba Cloud · 1M context
0.75× Alibaba Cloud's list price on a closed model, and nobody says why. Listed, never ranked.
$0.11 / $0.38in / out per M
0.78×of list price
- Qwen3.8 Max
Alibaba Cloud · 1M context
0.82× Alibaba Cloud's list price on a closed model, and nobody says why. Listed, never ranked.
$1.65 / $4.95in / out per M
0.83×of list price
Frequently asked
- What is DeepInfra?
- Serves open-weight models on its own GPUs, often at fp8 or fp4, and publishes the precision per model. It sells 106 of the models we track, and trades as Deep Infra, Inc., based in US.
- Is DeepInfra cheaper than buying from the maker?
- 84 of its 180 prices sit below the list price for the model, and 79 of those have a published reason behind them, so they can lead a ranking. 3 are listed and never ranked.
- Does DeepInfra train on what you send it?
- DeepInfra does not say whether it trains on what you send it. It does not say how long it keeps requests.
- How current are these DeepInfra prices?
- They come from its own public catalog, last read 16 Sept 2026. Prices exclude tax.