Free LLM API tiers

Free use of some Gemini models through AI Studio, with no billing set up.

• Free models include Gemini 3.8 Flash, 3.6 Flash, 3.5 Flash-Lite and 2.5 Pro; Gemini 4 Argon is not free
• Requests and tokens per minute and per day: not published; each project sees its own limits in AI Studio
• Daily quotas reset at midnight Pacific time, per project
• Payment method needed: No

Your data. Free-tier prompts and responses are used to improve Google products and may be read by human reviewers, except in the EEA, Switzerland and the UK.

Terms. Listed regions only, and 18 or older. Apps offered to users in the EEA, Switzerland or the UK must use the paid tier.

Read 2 Nov 2026 from Gemini API rate limits, Gemini API billing, Gemini API pricing · Google models on BenchLeader

A Free mode on La Plateforme for testing and prototyping, with no card.

• Capped on requests per second, tokens per minute and tokens per month; the numbers are shown only on the console's Limits page
• Payment method needed: No

Your data. Free-mode input and output may be used to train Mistral's models unless you opt out in the admin panel.

Read 2 Nov 2026 from Activate and generate an API key, Usage limits · Mistral AI models on BenchLeader

A Free plan with published limits per model.

• gpt-oss-120b: 30 requests a minute, 1,000 a day, 8,000 tokens a minute, 200,000 a day
• Qwen3.8 27B: the same 30 / 1,000 / 8,000 / 200,000
• Limits apply per model, per organisation; Groq says exceptions are possible
• Payment method needed: No

Your data. Groq says it does not retain inference data by default.

Read 2 Nov 2026 from Rate limits (Free plan tab), Billing FAQ

A trial key for every new account, for every Cohere model.

• 1,000 API calls a month
• Chat models: 20 requests a minute; Rerank: 10 a minute; Embed: 2,000 inputs a minute
• Payment method needed: No

Terms. Trial keys may not be used for production or commercial purposes.

Read 2 Nov 2026 from Rate limits, Pricing · Cohere models on BenchLeader

10,000 Neurons a day free, Cloudflare's compute unit, across the model catalogue.

• 10,000 Neurons a day; each model's token-to-Neuron rate is on the pricing page
• Text generation: 300 requests a minute by default
• Some models need a paid plan, among them Kimi K2.6, GLM 5.3 and DeepSeek V4 Pro
• Payment method needed: Not stated

Read 2 Nov 2026 from Workers AI pricing, Workers AI limits

Free use of hosted open models until you link a payment method.

• 20 requests a minute, 20 a day and 200,000 tokens a day, per model (DeepSeek-V3.1, Llama 3.3 70B, gpt-oss-120b; preview models the same)
• Payment method needed: No

Terms. Linking a payment method moves the account to the paid Developer tier.

Read 2 Nov 2026 from Rate limits

Hosted models on build.nvidia.com for prototyping, with a free developer account.

• Rate limits: not published
• Payment method needed: Not stated

Terms. NVIDIA Developer Program members only; prototyping, research and testing, not production, which needs an NVIDIA AI Enterprise licence.

Read 2 Nov 2026 from NVIDIA NIM, NIM API docs · NVIDIA models on BenchLeader

Models whose ID ends in :free cost nothing per token, within account-wide limits.

• 20 requests a minute across :free models
• 50 requests a day until the account has bought 10 credits in total, then 1,000 a day
• Payment method needed: No

Your data. Depends on the provider serving the model; account settings control whether free models may be routed to providers that train on prompts.

Read 2 Nov 2026 from API limits, FAQ

A small free credit for new users; the amount is not published.

• Credit amount and expiry: not published
• New organisations may start below the standard rate limits; the numbers are not published
• Payment method needed: Not stated

Read 2 Nov 2026 from Pricing, How API usage is billed, Rate limits · Anthropic models on BenchLeader

$5 of credit for a new account with a verified payment method, valid 30 days.

• $5, expiring 30 days after it is granted
• gpt-oss-120b on the trial: 5 requests a minute, 30,000 uncached tokens a minute, 1M tokens a day
• Payment method needed: Yes

Terms. No permanent free tier: after the credit, pay-as-you-go.

Read 2 Nov 2026 from Rate limits

A $1 starter credit; after that, prepaid credits.

• $1; expiry not published
• 10 requests a minute with no payment method or no credit left
• Payment method needed: No

Read 2 Nov 2026 from When the $1 credit runs out, Account quotas · Fireworks AI models on BenchLeader

Prepaid credits before the first call; only the Moderation endpoint is free.

• Minimum credit purchase $5; credits expire after a year
• The lowest usage tier is capped at $100 a month, in allowed countries
• Payment method needed: Yes

Read 2 Nov 2026 from Prepaid API billing, Rate limits and usage tiers · OpenAI models on BenchLeader

Prepaid top-ups; no standing free allowance is published.

• The billing rules mention a granted balance but give no amount; the minimum top-up is not published
• Payment method needed: Not stated

Read 2 Nov 2026 from Models and pricing · DeepSeek models on BenchLeader

Credits are added before the first call; no free credit is published.

• Accounts start at the lowest rate-limit tier; per-model limits are shown in the console
• Payment method needed: Not stated

Read 2 Nov 2026 from Quickstart, Rate limits · SpaceXAI models on BenchLeader

No free trial: access starts with a $15 credit purchase.

• $15 minimum first purchase, $5 after that
• Payment method needed: Yes

Read 2 Nov 2026 from Billing and credits · Together AI models on BenchLeader

No credit on a free account; paid PRO, Team and Enterprise plans include $2 a month.

• Free accounts: no monthly credit
• PRO: $2 a month; Team and Enterprise: $2 per seat a month
• Payment method needed: Yes

Read 2 Nov 2026 from Inference Providers pricing · Hugging Face models on BenchLeader