Skip to content
trending
// cheapest

Free LLM APIs on OpenRouter (Nov 2026)

Models with a free variant on OpenRouter, sorted by context window. Free variants are rate-limited and may be retired or changed without notice, so treat them as a way to prototype, not as production capacity.

The largest free window today belongs to Nemotron 3.5 Super VL (free): 1,000,000 tokens. 7 free models in total.

Data updated 2026-11-12 - 7 models - source: public OpenRouter model list, refreshed daily.

#ModelVendorContextMax outFeatures
1Nemotron 3.5 Super VL (free)NVIDIA1M131Kreasoning, tools, vision, video
2Nemotron 3.5 Lightning (free)NVIDIA1M66Kreasoning, tools
3Nemotron 3 Ultra (free)NVIDIA1M66Kreasoning, tools
4Gemma 4 26B A4B (free)Google262K33Kreasoning, tools, vision, video
5Gemma 4 31B (free)Google262K33Kreasoning, tools, vision, video
6North Mini Code (free)Cohere256K64Kreasoning, tools
7Nemotron 3 Nano Omni (free)NVIDIA256K66Kreasoning, tools, vision, audio, video

Context and max output in tokens (K = thousand, M = million).

Other lists

How we rank

The list includes text models from established vendors that OpenRouter added in the last 12 months, with at least a 32,000-token window. Batch, preview, experimental and image or audio generation variants are left out; batch prices appear in their own column on vendor pages. Prices are the current per-token list rates; for open-weight models served by several hosts, OpenRouter shows the price of the provider it currently routes to, which is why some of them move often. Need a full cost estimate? Use the AI cost calculator or browse the full model atlas.

Frequently asked questions

Are there free LLM APIs?

Yes. As of 12 Nov 2026, OpenRouter lists 7 free model variants from established vendors; Nemotron 3.5 Super VL (free) has the largest window among them at 1,000,000 tokens. Free variants are rate-limited and can change or disappear without notice.

How is the ranking calculated?

Models are sorted by the maximum number of input tokens they accept; ties are broken by the blended price (three input tokens for every output token).

How often is this list updated?

Every day. Prices, context windows and features come from the public OpenRouter model list, refreshed each morning; the date at the top of the page shows the last update.

// follow

New tests and price changes, straight to your feed reader

Free RSS, no account and no email needed.

Follow Quantslant →
// site map
Quantslant in numbers 1,429posts in archive25news from the last 7 days18video models in the observatory
🎬 AI Video Benchmarks: Higgsfield DoP vs OpenAI Sora • Higgsfield vs Runway Gen-3 • AI Video Credit Calculator
× ‹ enlarge ›