KeyForge

Every model, free-first, one key

Bring your own provider keys and route them all through one OpenAI-compatible endpoint. The counts below are computed straight from the live gateway registry , we only claim what we can actually route to today.

18
BYOK providers wired in
52
Catalog models
8
With a standing free tier
13
Free tier or free credits

We don’t advertise aggregator-style totals (some catalogs claim “169+ providers” by bundling OAuth IDEs, search and speech APIs). For the wider free-tier landscape, see the community-maintained free-llm-api-resources list.

Google Gemini

Free tier
US No-training default

Free tier via Google AI Studio (rate-limited; may train on prompts outside the EU/UK/CH).

Gemini 2.0 FlashGemini 1.5 ProGemini 1.5 Flash
Get a key

Groq

Free tier
US No-training default

Generous free tier, no card; per-minute and daily token limits apply.

Llama 3.3 70BLlama 3.1 8B InstantGPT-OSS 120B
Get a key

OpenRouter

Free tier
Mixed / aggregator

Includes :free model variants plus pay-as-you-go for the rest.

OpenAI GPT-4o (via OpenRouter)OpenAI GPT-4o mini (via OpenRouter)Claude 3.5 Sonnet (via OpenRouter)Gemini 2.0 Flash (via OpenRouter)Llama 3.3 70B (via OpenRouter)DeepSeek V3 (via OpenRouter)
Get a key

Mistral

Free tier
EU No-training default

Free experiment tier available; that tier may train on your data.

Mistral LargeMistral Small
Get a key

Cohere

Free tier
Canada No-training default

Free rate-limited trial keys; production keys are paid.

Command R+Command R
Get a key

Cerebras

Free tier
US No-training default

Free tier, no card; ~1M tokens/day and an 8K context cap on the free plan.

Llama 3.3 70B (Cerebras)Llama 3.1 8B (Cerebras)GPT-OSS 120B (Cerebras)
Get a key

SambaNova

Free tier
US No-training default

Free developer tier, no card; rate-limited (per-minute request cap).

Llama 3.3 70B (SambaNova)Llama 3.1 8B (SambaNova)DeepSeek V3 (SambaNova)
Get a key

NVIDIA NIM

Free tier
US No-training default

Free via the NVIDIA Developer Program; ~40 requests/min on the free tier.

Llama 3.3 70B (NVIDIA)DeepSeek R1 (NVIDIA)Nemotron 70B (NVIDIA)
Get a key

Together AI

Free credits
US No-training default

Free sign-up credits, then pay-as-you-go.

Llama 3.3 70B Turbo (Together)DeepSeek V3 (Together)Qwen2.5 72B Turbo (Together)
Get a key

Fireworks AI

Free credits
US No-training default

Free sign-up credits, then pay-as-you-go.

Llama 3.3 70B (Fireworks)DeepSeek V3 (Fireworks)
Get a key

Perplexity

Free credits
US No-training default

Monthly API credit for Pro subscribers; otherwise pay-as-you-go.

SonarSonar ProSonar Reasoning Pro
Get a key

Nebius

Free credits
EU No-training default

Free sign-up credits, then pay-as-you-go; EU-hosted inference.

Llama 3.1 70B (Nebius)DeepSeek V3 (Nebius)Qwen2.5 72B (Nebius)
Get a key

Hyperbolic

Free credits
US No-training default

Free sign-up credits, then pay-as-you-go.

Llama 3.3 70B (Hyperbolic)DeepSeek V3 (Hyperbolic)Qwen2.5 72B (Hyperbolic)
Get a key

OpenAI

Pay-as-you-go
US No-training default
GPT-4oGPT-4o miniGPT-4.1GPT-4.1 mini
Get a key

Anthropic

Pay-as-you-go
US No-training default
Claude 3.5 SonnetClaude 3.5 HaikuClaude 3 Haiku
Get a key

DeepSeek

Pay-as-you-go
China
DeepSeek V4 ProDeepSeek V4 Flash
Get a key

xAI (Grok)

Pay-as-you-go
US No-training default
Grok 4.6
Get a key

DeepInfra

Pay-as-you-go
US No-training default

Pay-as-you-go at low per-token prices; no standing free tier.

Llama 3.3 70B (DeepInfra)DeepSeek V3 (DeepInfra)Qwen2.5 72B (DeepInfra)
Get a key

Route the free ones first, automatically

Set a key’s provider sort to price and its cost-tier ceiling, and the Auto Router sends routine calls to free and lowest-cost capable models before it ever reaches a premium one, with a deterministic response cache that bills nothing on a repeat hit.

Start Free How cost routing works