Skip to content
OpenRouter Free Tokens and Free AI Models - 15 APIs (2026)

OpenRouter Free Tokens and Free AI Models - 15 APIs (2026)

OpenRouter offers 15 free models for testing and development, with no credit card required.

Free token rules: New OpenRouter users receive a small free credit; models marked :free are not billed per token, but default to 50 requests per day, rising to 1,000 per day after a $10 top-up. There is no official “monthly free token refresh” promise; refer to your account page and rate-limit documentation.
The models below are completely free, but some are limited-time trials and may become paid at any time. Use them for development and testing.

Model List (15)

2026-11-21 update: Synced the latest free list from the API. Inkling 2 Small’s :free variant was retired and is now paid ($0.30/M input, $1.00/M output); Cohere North Mini Code was removed from OpenRouter. Both removed; total free models 17 → 15.
2026-11-15 update: Synced the latest free list from the API. LFM 2.5 2.6B was removed from OpenRouter; added Ling 3.2 Flash (1M context, text+image, 262K output) and Particle-1 (1M context, agentic coding, 131K output). Total free models 16 → 17.
2026-11-11 update: Synced the latest free list. Inkling, Inkling Small, and Nemotron 3.5 Content Safety became paid; Xing4.0 29B A4B was removed. Added LFM3-1.7B, ERNIE 5.5 Flash Preview, and Inkling 2 Small. Total free models remains 16.
2026-11-03 update: Poolside Laguna S 2.1 and Laguna XS 2.1 were removed from OpenRouter; no new free models. Total 19 → 17.
2026-10-29 update: Nemotron 3 Super 120B (free) free endpoint was retired; the paid version is still available. Total 20 → 19.
2026-10-23 update: MiniMax M3.1 Flash (free) was removed and became paid; added Recursive Flint 1.0 (40B MoE, 262K context). Total remains 20.
2026-10-21 update: Added MiniMax M3.1 Flash (free), 1M context, text/image/video input. Total 19 → 20.
2026-10-18 update: Synced the latest free list. Removed 6 retired/paid models (Qwen3.8 27B, Nex-N2.5 Pro/Mini, GLM 5.2, Ling 3.0 Flash Fin/Sante); added 2 (Xing4.0 29B A4B, Nemotron 3.5 Super VL).

🟦 Google (4)

ModelContextDescription
Gemma 4 31B262KGoogle flagship dense multimodal
Gemma 4 26B A4B262KMoE efficient architecture
Lyria 3 Pro Preview1MGoogle full-song generation
Lyria 3 Clip Preview1MGoogle 30-second music clips

🟪 NVIDIA (4)

ModelContextDescription
Nemotron 3 Ultra 550B1MNEW 550B MoE frontier reasoning
Nemotron 3.5 Super VL1MNEW 120B MoE omni-modal (text+image+video)
Nemotron 3 Nano Omni256KMultimodal, supports video input
Nemotron 3.5 Lightning1MNEW 3.5-generation high-speed reasoning

🟣 Recursive (1)

ModelContextDescription
Recursive Flint 1.0262KNEW 40B MoE for coding agents

🟢 OpenRouter (1)

ModelContextDescription
OpenRouter Free200KAutomatically selects a free model

🔵 Other Vendors (5)

ModelContextDescription
Ling 3.2 Flash1MNEW inclusionAI hybrid reasoning, text+image, 200 images/request
Particle-11MNEW Particle AI agentic coding, based on DeepSeek V4.1 Flash
ERNIE 5.5 Flash Preview1MBaidu 280B text model, Preview
LFM3-1.7B262KLiquid 1.7B lightweight agent model
Dots3-Note Preview512KDots 3 lightweight MoE

📌 Removed (2026-11-21): Inkling 2 Small — :free variant retired, now paid; Cohere North Mini Code — removed from OpenRouter 📌 Removed (2026-11-15): LFM 2.5 2.6B — removed from OpenRouter 📌 Removed (2026-11-11): Inkling, Inkling Small, Nemotron 3.5 Content Safety — became paid; Xing4.0 29B A4B — removed from OpenRouter 📌 Removed (2026-11-03): Poolside Laguna S 2.1, Laguna XS 2.1 — removed from OpenRouter 📌 Removed (2026-10-29): Nemotron 3 Super 120B — :free endpoint retired, now paid 📌 Removed (2026-10-23): MiniMax M3.1 Flash — :free removed, now paid 📌 Removed (2026-10-18): Qwen3.8 27B, Nex-N2.5 Pro/Mini, GLM 5.2, Ling 3.0 Flash Fin/Sante — removed or paid 📌 Removed (2026-09-25): MiniMax M3, MiniMax M2.7 — removed from OpenRouter 📌 Removed (2026-09-04): Llama 3.3 70B, Llama 3.2 3B, GPT-OSS 120B/20B, Qwen3 Coder, Qwen3 Next 80B, Hermes 3 405B, Dolphin Venice, LFM 1.2B ×2, Nemotron Nano ×3, Laguna M.1/XS.2 — no longer free or removed

Usage Example

# Call a free model
curl https://openrouter.ai/api/v1/chat/completions \
  -H "Authorization: Bearer $OPENROUTER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "nvidia/nemotron-3.5-super-vl:free",
    "messages": [{"role": "user", "content": "Hello!"}]
  }'

Notes

  • 2026-11-21 update: Inkling 2 Small became paid ($0.30/$1.00 per million tokens); Cohere North Mini Code was removed. Total 17 → 15.
  • 2026-11-15 update: LFM 2.5 2.6B removed; added Ling 3.2 Flash and Particle-1. Total 16 → 17.
  • 2026-11-11 update: Inkling, Inkling Small, and Content Safety became paid; Xing4.0 removed. Total remains 16; added LFM3, ERNIE 5.5, and Inkling 2 Small.
  • 2026-11-03 update: Poolside Laguna S 2.1 and Laguna XS 2.1 removed. Total 19 → 17.
  • 2026-10-29 update: Nemotron 3 Super 120B (free) endpoint retired. Total 20 → 19.
  • 2026-10-23 update: MiniMax M3.1 Flash (free) removed and became paid; added Recursive Flint 1.0. Total remains 20.
  • 2026-10-21 update: Added MiniMax M3.1 Flash (free). Total 19 → 20.
  • 2026-10-18 update: Synced the latest free list (19). Removed 6: Qwen3.8 27B, Nex-N2.5 Pro/Mini, GLM 5.2, Ling 3.0 Flash Fin/Sante; added 2: Xing4.0 29B A4B and Nemotron 3.5 Super VL.
  • 2026-09-25 update: Synced the latest free list (23). Removed MiniMax M3/M2.7; added Qwen3.8 27B, Nex-N2.5 Pro/Mini, and Ling 3.0 Flash Sante (4 total).
  • 2026-09-04 update: Synced the latest free list (21). Removed 15 retired models including Llama 3.3/3.2, GPT-OSS, Qwen3 Coder, and Hermes 3 405B; added 11 including MiniMax M3/M2.7, GLM 5.2, Inkling, and Nemotron 3.5 Lightning.
  • 2026-07-02 update: Fixed 5 filenames (added the :free suffix); removed the retired OpenRouter Owl Alpha.
  • 2026-06-27 update: Removed the expired Nex AGI Nex-N2-Pro.
  • Free model naming: most free models use the :free suffix; a few, such as Lyria 3 and OpenRouter Free, are billed at $0/$0.
  • Limited-time trial models: some free models may become paid at any time; check regularly.
  • Data usage: inputs and outputs of free models may be used for training.
  • Production: use free models for development and testing; use paid models in production.

Related Links

Last updated on · Source: OpenRouter official documentation and pricing pages. Quotas and prices may change at any time.