Free LLM APIs
17 models currently have a completely free API tier. The most capable is North Mini Code (free) (19.8 Intelligence Index), followed by Nemotron 3 Nano Omni (free). All of them are callable through a single OpenRouter key at $0 per token.
Every model with a genuinely free API tier, ranked by intelligence. These are real frontier-lab and open-weight models served at $0 per token (rate-limited) — not trials that expire. Looking for machine-readable benchmark data instead? modelgrep's own free JSON API needs no key at all.
- 1
north-mini-code:freeReasoningTools256K ctx19.8Intelligence - 2
nemotron-3-nano-omni-30b-a3b-reasoning:freeReasoningToolsVision+1256K ctx14.9Intelligence - 3
gpt-oss-20b:freeReasoningToolsJSON131K ctx14.9Intelligence - 4Iling-3.0-flash:freeReasoningTools262K ctx—Intelligence
- 5Plaguna-s-2.1:freeReasoningTools262K ctx—Intelligence
- 6Plaguna-xs-2.1:freeReasoningTools262K ctx—Intelligence
- 7nemotron-3.5-content-safety:freeReasoningVision128K ctx—Intelligence
- 8nemotron-3-ultra-550b-a55b:freeReasoningTools1M ctx—Intelligence
- 9Plaguna-m.1:freeReasoningTools262K ctx—Intelligence
- 10
gemma-4-26b-a4b-it:freeReasoningToolsJSON+1262K ctx—Intelligence - 11gemma-4-31b-it:freeReasoningToolsJSON+1262K ctx—Intelligence
- 12lyria-3-pro-previewJSONVision1.0M ctx—Intelligence
- 13lyria-3-clip-previewJSONVision1.0M ctx—Intelligence
- 14nemotron-3-super-120b-a12b:freeReasoningToolsJSON262K ctx—Intelligence
- 15nemotron-3-nano-30b-a3b:freeReasoningTools256K ctx—Intelligence
- 16nemotron-nano-12b-v2-vl:freeReasoningToolsVision128K ctx—Intelligence
- 17nemotron-nano-9b-v2:freeReasoningToolsJSON128K ctx—Intelligence
Frequently asked
Is there a free LLM API?
Yes — 17 models on this page charge $0 per token via OpenRouter's free tier (model IDs ending in ":free"). Rate limits apply, so they suit prototyping and low-volume production, not high-traffic apps.
What is the best free LLM API?
North Mini Code (free) is the most capable free model right now, scoring 19.8 on the Intelligence Index — see the full ranking below for speed and context tradeoffs.
How do I call a free model?
Create a free OpenRouter API key, then call the OpenAI-compatible /chat/completions endpoint with the model's ID (e.g. the ":free" variant shown on its page). No credit card is required for free-tier models.
Is there a free API for model metadata too?
Yes — modelgrep's own free, no-key JSON API serves benchmarks, pricing and speed data for all 300+ models at modelgrep.com/api.