InferIndex
Language, English, EN
Account

ModelsMicrosoftPhi 4

Phi 4

microsoft/phi-416k context Compare →
MedianThe median blended price among comparable offers: one price per provider (its cheapest route), standard tier, current prices only. Priority tiers and prices converted from points are not counted.
$0.088
Most expensiveThe highest blended price among comparable offers: one price per provider (its cheapest route), standard tier, current prices only. Priority tiers and prices converted from points are not counted.
$0.088
Offers
9
Providers
4

Cheapest right now

DeepInfra

The provider's published policy: prompts are not used for training.

directYou buy access straight from the company that runs the model on its servers. · bf16 · 16k

$0.088 blended / 1M tokens

No change7D
in
$0.07
out
$0.14
Cached input
–
Go to offer Go to offer at DeepInfra (opens in a new tab)at DeepInfra

Public leaderboards

  • Text Arena, style controlThe series the source publishes by default for text since 16 May 2025: ranks corrected for the style of the answers (style control). #317 of 419Best rank in the source's rank spread, 317th to 325th, given the confidence intervals of all ranked models. 323rd by score. 1,256 1,252 to 1,261Arena Score · 24,126 votes

Source: Arena (opens in a new tab) · Data under the CC BY 4.0 license (creativecommons.org/licenses/by/4.0) · CC BY 4.0 (opens in a new tab) · scores and ranks as published on 28 Oct 2026; the presentation is ours

Phi 4 costs from $0.088 per million tokens (at a 3:1 input:output mix): the lowest current price, at 2 providers, checked on ; 9 offers are listed.

Price historyThe cheapest offer each day, against the lab's official list price when it publishes one. Each day at 00:00 UTC, the offer the API would have selected as the cheapest, with the same exclusions as /cheapest. Blended price, promotions included.

Cheapest offer and official list price, by day, 30 days
DayCheapest offerOfficial list price
2026-10-02$0.088, DeepInfra–
2026-10-03$0.088, DeepInfra–
2026-10-04$0.088, DeepInfra–
2026-10-05$0.088, DeepInfra–
2026-10-06$0.088, DeepInfra–
2026-10-07$0.088, DeepInfra–
2026-10-08$0.088, DeepInfra–
2026-10-09$0.088, DeepInfra–
2026-10-10$0.088, DeepInfra–
2026-10-11$0.088, DeepInfra–
2026-10-12$0.088, DeepInfra–
2026-10-13$0.088, DeepInfra–
2026-10-14$0.088, DeepInfra–
2026-10-15$0.088, DeepInfra–
2026-10-16$0.088, DeepInfra–
2026-10-17$0.088, DeepInfra–
2026-10-18$0.088, DeepInfra–
2026-10-19$0.088, DeepInfra–
2026-10-20$0.088, DeepInfra–
2026-10-21$0.088, DeepInfra–
2026-10-22$0.088, DeepInfra–
2026-10-23$0.088, DeepInfra–
2026-10-24$0.088, DeepInfra–
2026-10-25$0.088, DeepInfra–
2026-10-26$0.088, DeepInfra–
2026-10-27$0.088, DeepInfra–
2026-10-28$0.088, DeepInfra–
2026-10-29$0.088, DeepInfra–
2026-10-30$0.088, DeepInfra–
2026-10-31$0.088, DeepInfra–

Cheapest offer

Offers

9 of 9 offers

Offers for Phi 4, one row per offer, sorted by blended price.
RankProvider · viaYou buy access through an aggregator: one account and one API that forward your requests to other providers. "Via" names the aggregator; the price is the one it charges you. When "via" says "hosted API", no aggregator is involved: the model belongs to a lab and another provider sells it.Input $/1MOutput $/1MBlended $/1MA single price that mixes input and output: 3 input tokens for 1 output token, per million tokens. It makes offers easy to rank; your real cost depends on your own mix.PrecisionHow finely the model's numbers are stored on the provider's servers. Lower precision (fp8, fp4) is cheaper to run and can slightly change answer quality; "unknown" means the provider doesn't say. ContextThe maximum amount of text, counted in tokens, the model can take in and write in a single request. Past this limit, the request is refused or cut.ConditionsWhat a provider says about how it handles your data and who can sign up: training on prompts, processing region, sign-up requirements. Each comes from the provider's published documents; when nothing is published, it stays unknown.SignalsOffer
1 DeepInfra 0.07 0.14 0.088 bf16 16k United States
Go to offer at DeepInfra (opens in a new tab)
2 Featherless 0.07 0.14 0.088 – 33k
Go to offer at Featherless (opens in a new tab)
3 DeepInfravia Eden AI 0.07 0.14 0.088 bf16 16k United States Unverified. No public page for this offer: listed, not designated cheapest
: Eden AI (opens in a new tab). No public price page for this model
4 DeepInfravia Hugging Face 0.07 0.14 0.088 bf16 16k United States
See pricing at Hugging Face (opens in a new tab)
5 DeepInfravia Kilo Gateway 0.07 0.14 0.088 bf16 16k United States Unverified. The source page shows one price per model, not this seller's: listed, not designated cheapest
Go to offer at Kilo Gateway (opens in a new tab)
6 Glama 0.07 0.14 0.088 – 16k Unverified. No public page for this offer: listed, not designated cheapest
: Glama (opens in a new tab). No public price page for this model
7 Martian 0.07 0.14 0.088 – – worldwide Unverified. Neither seller nor price found on its page for 72 hours: listed, not designated cheapest
See pricing at Martian (opens in a new tab)
8 DeepInfravia Requesty 0.07 0.14 0.088 bf16 16k United States Fee 5%: Requesty adds a 5% markup (read on requesty.ai/pricing on Oct 30). Offers are ranked with this fee included.
Go to offer at Requesty (opens in a new tab)
9 DeepInfravia OpenRouter 0.07 0.14 0.088 bf16 16k United States Fee 5.5%: OpenRouter adds a 5.5% fee when you buy credits (read on openrouter.ai/pricing on Oct 30). Offers are ranked with this fee included.
Go to offer at OpenRouter (opens in a new tab)

No training on promptsPrompts may be used for trainingTraining terms not published

backend not named. The router bills this price itself without naming who serves the model.

4 flex or batch offers hidden (lower priority, variable latency).

Blended = (3 × input + output) / 4, per million tokens. Prices in USD, converted at the ECB rate when the provider bills in another currency.

Method

Monthly cost for your workload

Computed by the API for the three cheapest offers, per 30-day month.
  1. Featherless$6.30per month

Host it yourself instead? See the break-even

Compare Phi 4 with

Related models

More from Microsoft

All Microsoft models

Same index family

Use this data

The index is queryable. The API is free to use.
Free API
curl https://api.inferindex.dev/cheapest?model=microsoft/phi-4
MCP server
claude mcp add --transport http inferindex https://mcp.inferindex.dev/mcp

Also works in Cursor and Codex: see the MCP page.