Buyer's guide

Best Free LLM API: four free and trial options

Choose Groq for recurring free GPT-OSS 120B access at 30 RPM, or Cerebras for $5 in trial credits valid for 30 days and a Developer-tier advertised model speed of ~3,000 tokens/s.

Updated · 10 sources · Methodology

NeedPickWhy
Lowest listed token costGroq / OpenRouter$0 input / $0 output per 1M tokens within Groq Free and OpenRouter free-variant quotas.129
Highest published model speedCerebras~3,000 vs ~500 tokens/s for GPT-OSS 120B on Cerebras’s Developer listing and Groq’s model listing, respectively (vendor-reported, 2026).510
Workers integrationCloudflare Workers AI10,000 Neurons/day shared across Workers AI models in the included allowance.7
Time-limited trialCerebras$5 in trial credits for 30 days to try shared inference; a verified payment method is required.4

Capabilities and serving conditions

Free allowances and trials have different limits

Swipe to see all columns

ProviderReference API or modelIncluded allowancePublished request ceiling
Groq1GPT-OSS 120B
Free tier
1K requests/day and 200K tokens/day, with an 8K tokens/minute limit.130 RPM
GPT-OSS 120B Free plan
OpenRouter3:free variants
Free tier
50 free requests/day by default; eligible accounts receive 1,000/day.320 RPM
Account-wide free-variant limit
Cloudflare Workers AI8@cf/openai/gpt-oss-120b
Free tier
10,000 Neurons/day shared across models; this is a compute allowance, not a token or request count.7300 RPM
Default text-generation limit
Cerebras4GPT-OSS 120B
Time-limited trial
A one-time $5 credit balance that expires after 30 days.45 RPM
GPT-OSS 120B Free Trial

Published GPT-OSS 120B speed figures

Swipe to see all columns

DimensionGroq10Cerebras5
Published output speed (tokens/s) ~500
vendor-reported (2026)10
~3,000
vendor-reported (2026)5
Published scopeModel listingDeveloper tier model listing

Pricing

Pricing

Swipe to see all columns

ProviderInput / 1M tokensOutput / 1M tokensHow the allowance applies
Groq1$0.00$0.00Zero token charges within the selected free model’s quotas.
OpenRouter29$0.00$0.00Zero token charges for an available :free variant within account quotas.
Cloudflare Workers AI67$0.35$0.75The shared daily Neurons allowance covers eligible usage; the rates describe metered usage before that allowance.
Cerebras45$0.35$0.75Metered usage reduces trial credits; new cash payment is not required while valid trial credit remains.
Metered value: 1,000 input + 100 billed output tokens, before allowances
Groq$0.00
OpenRouter$0.00
Cloudflare Workers AI$0.000425
Cerebras$0.000425

Options, strengths and limitations

What each option is for

  1. Groq

    A direct API for experimenting with GPT-OSS 120B under a recurring free quota.

    Input / 1M tokens
    $0.00
    Output / 1M tokens
    $0.00
    Published request ceiling
    30 RPM

    Strengths

    • Minute and daily quotas are published for the selected model, so a prototype can budget requests and tokens separately.

    Limitations

    • The first request or token limit reached stops further usage; spare daily capacity does not raise the minute limit.

    Sources:1

  2. OpenRouter

    A model-discovery option for applications that switch among available free variants.

    Input / 1M tokens
    $0.00
    Output / 1M tokens
    $0.00
    Published request ceiling
    20 RPM

    Strengths

    • One API integration accesses the available free-model catalog without buying credits for the base free allowance.

    Limitations

    • Free variants have account-wide quotas and no contractual SLA; the catalog can change.

    Sources:239

  3. Cloudflare Workers AI

    An option for applications that run inference from Workers or through the Workers AI REST API.

    Input / 1M tokens
    $0.35
    Output / 1M tokens
    $0.75
    Published request ceiling
    300 RPM

    Strengths

    • The included daily compute budget can be used across Workers AI models instead of maintaining a separate trial credit balance.

    Limitations

    • The Free plan stops serving once the daily compute allowance is exhausted; Paid plans bill usage above the allowance.

    Sources:678

  4. Cerebras

    A bounded trial of shared inference before moving to purchased credits.

    Input / 1M tokens
    $0.35
    Output / 1M tokens
    $0.75
    Published request ceiling
    5 RPM

    Strengths

    • Trial credit can be used across public Shared Inference models, rather than being reserved for one demo model.

    Limitations

    • A verified payment method is required to activate the trial; API access stops when the trial expires or its credit runs out.

    Sources:45

Buyer's guide

How the options are ordered

  • Recurring free-access programs precede the time-limited trial.
  • The order groups direct model access, free-model selection and Workers integration; it is not an output-speed ranking.

Methodology

Methodology and sources

Groq and Cerebras reference GPT-OSS 120B; OpenRouter uses available :free variants, and Cloudflare uses @cf/openai/gpt-oss-120b. The speed row compares only Groq’s model listing with Cerebras’s Developer listing: vendor-reported figures, not measured free-tier performance or a common test setup. Cloudflare’s free eligibility follows its allowance rules and paid-only model list. Neurons are shared compute units; the published RPM ceilings apply to the plan/model scopes in the matrix. OpenRouter’s higher daily allowance uses lifetime purchased credits: the published 10-credit threshold currently has a one-credit grace, granting it from 9 credits.

The chart values 1,000 input and 100 billed output tokens before free allowances or trial credits; billed output includes reasoning. They are metered usage values, not invoices or guaranteed request counts. Taxes, retries and discounts are omitted. Novita AI compiled this free-access guide from public sources without running an account workload.

  1. Groq · Free Plan Rate Limits console.groq.com · Checked 2026-10-16
  2. OpenRouter · Free Plan and Billing openrouter.ai · Checked 2026-10-16
  3. OpenRouter · Free-Variant Limits openrouter.ai · Checked 2026-10-16
  4. Cerebras · Trial Terms and Limits inference-docs.cerebras.ai · Checked 2026-10-16
  5. Cerebras · Developer Token Pricing www.cerebras.ai · Checked 2026-10-16
  6. Cloudflare · GPT-OSS 120B developers.cloudflare.com · Checked 2026-10-16
  7. Cloudflare · Workers AI Pricing developers.cloudflare.com · Checked 2026-10-16
  8. Cloudflare · Workers AI Limits developers.cloudflare.com · Checked 2026-10-16
  9. OpenRouter · Free Models Router openrouter.ai · Checked 2026-10-16
  10. Groq · Published Model Speeds console.groq.com · Checked 2026-10-16

Build with Novita Model APIs

Create an API key and connect your application.

Everything you need to build production AI.

200+ models, on-demand GPUs, and secure agent runtimes — unified under one API. Free to start, scales as you grow.