First Deals vertical

Free AI inference without the hidden catches

Compare useful capacity, model access, account requirements, rate limits, and post-allowance behavior. Permanent free tiers stay separate from temporary credits and experiments.

Read the full comparison guide How verification works

Latest source check: 2026-11-04 UTC.

Capacity over marketing

“Free” means little without daily tokens, requests, model access, and reset behavior.

Prototype vs production

Experimental endpoints and rotating models are useful, but clearly labeled as unstable.

Official evidence

Every changing claim links to the provider and carries a verification date.

Clear categories

Different kinds of “free” are not interchangeable

Filter verified offers

TypeUse caseRegionAudienceValueExpiry

Verified offers

Showing 5 of 5; commission never affects ranking.

Free Tier

OpenRouter

Verified

Free-model router and zero-cost model variants

The zero-price listing adds HeyGen Voice Turbo (dated October 26), Recursive Flint, and NVIDIA Nemotron 3.5 Embed 1B and 8B (dated October 21), all absent from the master. Voice Turbo supports 39 languages and up to 5,000 characters per request and is marked as going away November 30; Flint lists 262K context and the embedding models return 2048- and 4096-dimensional vectors respectively.

Value: $0/M input and $0/M output tokens on four listed hosted models; Free plan lists 50 requests/day
Who qualifies: OpenRouter users with access to the relevant API models, subject to account request caps and model availability; own HeyGen key needed for voice clones from the user's account
Expiry: HeyGen Voice Turbo: going away November 30, 2026; other listed models: not stated
Checked: Nov 4, 2026

PrototypingModel TestingOpenAI-compatible

Details Open offer

Official source

Smartoolbox does not currently earn a commission from this offer.

Credits

Cerebras

Verified

Cerebras clarifies $5 trial: payment method required, expires after 30 days

Cerebras pricing now explicitly requires a valid payment method for its one-time $5 promotional credit, which expires 30 days after activation. API and Playground access pause when credit is exhausted or expires unless the user separately buys PayGo credits; there is no automatic paid enrollment.

Value: $5 one-time inference credit valid for 30 days after activation
Who qualifies: Users who add a valid payment method; promotional credits may be withheld or revoked for fraud, abuse, or duplicate accounts; intended for development, evaluation, and experimentation, not production
Expiry: 30 days after activation
Checked: Nov 4, 2026

AI Inference

Open offer

Official source

Smartoolbox does not currently earn a commission from this offer.

Free Tier

Google

Verified

Gemini 4 Argon launch pricing halves input and output rates through November 30

Official Gemini API pricing lists Gemini 4 Argon input at $2/M and output including thinking at $10/M through November 30, 2026, rising to $4/M and $20/M from December 1. This is paid hosted inference, not a free-tier offer; the model is explicitly unavailable on the Free tier.

Value: $2/M input and $10/M output through November 30, versus $4/M and $20/M starting December 1
Who qualifies: Gemini Developer API paid-tier users with access to gemini-4-argon; not available on the Free tier
Expiry: November 30, 2026; higher rates start December 1, 2026
Checked: Nov 4, 2026

AI Inference

Open offer

Official source

Smartoolbox does not currently earn a commission from this offer.

Free Tier

Nous Research

Verified

Nous Portal free catalog rotates to five zero-priced hosted models

The authenticated catalog snapshot lists five zero-priced hosted models: meituan/longcat-2.5:free, reflection/beam-1.0-501b-a23b:free, stealth/crouton-alpha, stepfun/step-3.7-flash:free, and upstage/solar-mini4:free. Compared with the October 29 six-model snapshot, LongCat 2.5 and Crouton Alpha appear while Ornith and both Poolside variants are absent; availability rotates.

Value: Five hosted models at $0 prompt and $0 completion pricing at collection time
Who qualifies: Nous Portal account holders with catalog access; numeric quotas and production guarantees not stated in the supplied snapshot
Expiry: not stated; catalog can rotate
Checked: Nov 4, 2026

AI Inference

Open offer

Official source

Smartoolbox does not currently earn a commission from this offer.

Temporary Promotion

Replicate

Verified

Replicate advertises 50% off Minimax H3 Lite and Highspeed until year end

Replicate's pricing-page banner advertises a 50% promotion for Minimax H3 Lite and Highspeed until the end of the year. This differs from the previously tracked Wan 3.0 discount; exact promotional unit prices and model endpoint URLs are not provided in the excerpt.

Value: 50% promotional pricing on Minimax H3 Lite and Highspeed
Who qualifies: Replicate users running the named models while promotional pricing applies; no additional eligibility stated
Expiry: until end of year, per banner collected November 4, 2026; exact cutoff time not stated
Checked: Nov 4, 2026

AI Inference

Open offer

Official source

Smartoolbox does not currently earn a commission from this offer.

Need a durable comparison?

The maintained guide compares model access, useful capacity, card requirements, data treatment, regions, and what happens when the free allowance ends.

Open the 2026 free inference guide