Ranking · November 2026Updated November 18, 2026

Private AI APIs, ranked by what they keep

Silk Compute ranks first of ten private AI APIs on how long they keep your prompts.

Silk Compute does not store your prompts or model outputs.

23 policy pages quotedHow the ranking works
The short answerDefault retention
Up to 24 hoursUnfil.AI
55 daysGemini API
No period publishedMistral API
See the timeline
Not stored
3of 10 APIs

The three are Silk Compute, abliteration.ai and Featherless.

RetentionSourceRetention timeline
Shortest stated period
24hours

Unfil.AI deletes temporary data within 24 hours.

RetentionSourceunfil.ai
Most common default
30days

xAI, OpenAI, Anthropic and Morph keep prompts for up to 30 days.

RetentionSourceRetention timeline
Longest default
55days

Google keeps Gemini API prompts for 55 days.

RetentionSourceaistudio.google.com
Kept if flagged
2years

Anthropic keeps flagged chats up to 2 years.

Flagged contentSourceprivacy.claude.com
Policy pages quoted
23pages

Each quote links to the provider's own page.

SourcesSourceMethod

The most private AI APIs

The order follows default retention, then what keeping nothing takes.

Method
#APIPrompts kept by defaultTraining on API promptsKeeping nothing takesPay with crypto
1Silk ComputeOpen-weight modelsZero data retentionPrompts and responses are never stored.A verified emailNo approval step or contractMonero, Bitcoin, Ether, USDT, USDC
2abliteration.aiOpen-weight modelsNever stored by defaultWeb search and fetch excludedNever used to trainNothing, on by defaultNot listed
3FeatherlessOpen-weight modelsNot loggedNot statedPrompts are not storedNothing, on every planNot listed
4Unfil.AIOpen-weight modelsUp to 24 hoursTemporary processing dataNot used for trainingNo option statedYes
5xAI APIGrok models30 daysAbuse and misuse auditsNot without permissionA console settingTeam-wide, not yet on for every teamNot listed
6OpenAI APIGPT modelsUp to 30 daysAbuse-monitoring logsNot unless you opt inPrior approvalPlus additional requirementsNot listed
7Claude APIClaude modelsWithin 30 daysUp to 2 years if flaggedNot by defaultFeedback you send is an exceptionA sales requestCovered Models keep 30 daysNot listed
8MorphOpen-weight modelsUp to 30 daysFree and paid tiersNot on any tierAn enterprise addendumUSDC
9Gemini APIGemini models55 daysPrompts, context and outputYes on the unpaid tierNo on the paid tierSettings, or Vertex AIVertex AI for a guaranteeNot listed
10Mistral APIMistral modelsNo period publishedMay train on Free modePay-as-you-go can opt outA written requestAt Mistral's discretionNot listed
privateaiapi.com is published by the team behind Silk Compute, one of the providers ranked here.Read November 18, 2026
PRIVATE AI APICC-BY
Retention

How long each API keeps a prompt

The hatched bar shows what Anthropic keeps when a chat is flagged.

All sources
Default retention of API prompts and outputs
Not storedKept by defaultKept longer if flagged
Silk ComputeOpen-weight models
Zero data retention
abliteration.aiOpen-weight models
Never stored by default
FeatherlessOpen-weight models
Not logged
Unfil.AIOpen-weight models
Up to 24 hours
xAI APIGrok models
30 days
OpenAI APIGPT models
Up to 30 days
Claude APIClaude models
Within 30 daysUp to 2 years if flagged2 years if flagged
MorphOpen-weight models
Up to 30 days
Gemini APIGemini models
55 days
Mistral APIMistral models
No period published
Log scale, read November 18, 2026.
PRIVATE AI APICC-BYSources for every bar

What keeping nothing costs you

Nine of the ten offer a way to keep nothing, and what it takes differs.

How to get it from each lab
Nothing to request

Nothing

There is no form, tier or contract.

A setting or a request

A setting or approval

You switch it on, apply for it or justify it.

A contract or another product

A contract

It goes through sales, an enterprise addendum or another cloud product.

Open-weight model APIs, side by side

APIPrompts keptContextAPI formatsPayment
Silk ComputeZero data retention1,048,576 tokens on all six modelsChat Completions, Messages, Responses (stateless)Card or crypto. Plans from $10 to $200 a month, or credits from $5 that never expire.
abliteration.aiNever stored by default1M tokens on its large modelsChat Completions, Messages, ResponsesCard. Plans of $20, $50 or $200 a month, or prepaid from $5.
FeatherlessNot logged32K on its top abliterated modelChat CompletionsCard or wallets. A $25 or $50 plan, or pay as you go.
Unfil.AIUp to 24 hours128K on Hermes 3 405BChat CompletionsCard, wallets or crypto, pay as you go.
MorphUp to 30 days1M tokensChat Completions, Messages, ResponsesCard or USDC. Prepaid from $10, or a $200 plan.
Prices and plans from each provider's pricing page, read November 18, 2026.Head-to-head comparisons