Phi 4 pricing.
Prices checked
Phi 4 costs $0.07 in and $0.14 out per million tokens from OpenRouter; the cheapest price is $0.07 at Kilo Gateway, checked .
OpenRouter's own price
OpenRouter
$0.07 in · $0.14 out
90% below the typical price of the 392 text models we track.
Cheapest price
Kilo Gateway
$0.06 in · $0.11 out
20% less than OpenRouter, and the provider says why.
Every provider selling Phi 4
- 1
$0.06 in · $0.11 out
$0.07/M tokens - 2
$0.06 in · $0.11 out
$0.07/M tokens - 3
$0.07 in · $0.14 out
$0.09/M tokens - 4
$0.07 in · $0.14 out
$0.09/M tokens - 5
$0.07 in · $0.14 out
$0.09/M tokens - 6
$0.07 in · $0.14 out
$0.09/M tokens - 7
$0.07 in · $0.14 out
$0.09/M tokens - 8
$0.07 in · $0.15 out
$0.09/M tokens - 9
$0.13 in · $0.50 out · 128k context here
$0.22/M tokens - 10
$0.13 in · $0.50 out · 128k context here
$0.22/M tokens - 11
$0.13 in · $0.50 out
$0.22/M tokens
Not ranked
Below the maker's own price with nothing published to account for it, from a catalog we could not read reliably, or a free tier, which is a limit rather than a price. Why
$0.07 in · $0.14 out · 33k context here
$0.09/M tokens
What it costs for your work
A month at OpenRouter's own price
A month at Kilo Gateway, the cheapest
What you save
A page is about 600 tokens in and 60 out, a support reply 200 in and 300 out, an email 150 in and 250 out. Nothing else is counted: no cache discount, no batch rate, no free tier.
Run it yourself
On a server by the month
Microsoft publishes the weights, so you can run Phi 4 on a server you rent by the month. It needs about 12.2 GB of RAM on a CPU, and the cheapest plan we track with that much and a disk included is Leaseweb VPS 2 at $9.04 a month, streaming about 1.6–3.2 tok/s. Every plan that fits
On a GPU by the hour
The RTX A5000 holds Phi 4 in 24 GB of memory against the 12.2 GB the weights, the context and the runtime need. The cheapest one we track is RunPod's RTX A5000 · Community Cloud at $0.16/GPU-hr, which is $116.80 for the 730 hours in a month, checked 31 Oct. Every GPU price
How we count
Prices are per million tokens. The big number on each row is blended 3 in : 1 out, the way most work uses a model. A price below OpenRouter's own is listed as cheapest only when the provider says why. Microsoft trained Phi 4 and sells it on no catalog we read, so the price every row is measured against is OpenRouter's.
- Resells
- The provider sells another company's endpoint, at a margin of its own or at the maker's price.
- Runs it themselves
- The provider runs the open weights on hardware of its own, so the price is its own rather than a resale.
- Same as maker
- Sold by someone else at the price the maker charges.
- Costs more
- Sold by someone else for more than the maker charges.
- Not ranked
- Below the maker's price on a closed model with nothing to account for it, or from a catalog we could not read reliably. Listed, and never a pick.
Similar models
Models from other labs that do what Phi 4 does, cheapest first. Each percentage is against Phi 4's cheapest price, $0.07 per M tokens.
Frequently asked
- Who sells Phi 4 cheapest?
- Kilo Gateway, at $0.06 in and $0.11 out per million tokens, checked 31 Oct 2026. The provider sells another company's endpoint, at a margin of its own or at the maker's price.
- Is Phi 4 open weights?
- Yes. Microsoft publishes the weights, so anyone may host Phi 4, which is why 10 providers sell it and why their prices differ so much.
- How much context does Phi 4 hold?
- 16,384 tokens in one call, which is about 16k.
- How does Phi 4 compare with the rest of the market?
- At $0.09 per million tokens, OpenRouter's own price is 90% below the typical $0.88 across the 392 text models we track.