OpenRouter API Pricing: 21 Free Models, 10 Price Changes, Cheapest Paid Input at $0.017/M
OpenRouter lists 21 free models and 430 total models as of 2026-09-10, with 10 models showing price changes since yesterday and the cheapest paid input price at $0.017 per million tokens for IBM Granite 4.0 Micro.
Free model inventory
The free tier holds 21 models from seven providers. Google contributes four: Lyria 3 Pro Preview, Lyria 3 Clip Preview, Gemma 4 26B A4B, and Gemma 4 31B, each with 262,144 or 1,048,576 token context windows. NVIDIA provides five Nemotron variants — 3.5 Lightning, 3 Ultra, 3 Super, 3 Nano Omni, and 3.5 Content Safety — with context windows ranging from 128,000 to 1,000,000 tokens. Thinking Machines offers Inkling Small and Inkling at 1,048,576 tokens. Nex AGI lists Nex-N2.5-Pro and Nex-N2.5-Mini at 262,144 tokens. inclusionAI has Ling 3.0 Flash Sante and Ling 3.0 Flash Fin at 262,144 tokens. Poolside contributes Laguna S 2.1 and Laguna XS 2.1 at 262,144 tokens. Cohere adds North Mini Code at 256,000 tokens. OpenRouter's own Free Models Router sits at 200,000 tokens. Dots Studio provides Dots3-Note Preview at 512,000 tokens.
Model churn
One model left the catalog: Nous Hermes 4 70B from nousresearch. No new models were added. No models became free and none stopped being free compared to 2026-09-09.
Price changes
Ten models updated pricing. Z.ai GLM 5.3 Flash dropped input from $0.075 to $0.07 and output from $0.25 to $0.2333 per million tokens. Z.ai GLM Flash Latest mirrored the same change. Z.ai GLM 4.6 cut input from $0.55 to $0.43 and output from $2.20 to $1.75. DeepSeek V4 Pro 0813 raised input from $0.5795 to $0.66 and output from $1.7384 to $1.98. DeepSeek V4 Flash 0423 increased input from $0.0809 to $0.0886 and output from $0.1618 to $0.1772. DeepSeek V4 Pro 0423 lowered input from $0.9396 to $0.87 and output from $1.8792 to $1.74. Qwen Qwen3 235B A22B Instruct 2507 more than doubled input from $0.09 to $0.22 and output from $0.55 to $0.88. Qwen Qwen3 Next 80B A3B Instruct trimmed input from $0.10 to $0.09 with output unchanged at $1.10. IBM Granite 4.2 8B reduced input from $0.10 to $0.06 but raised output from $0.15 to $0.25. MoonshotAI Kimi Latest cut input from $2.55 to $2.40 and output from $12.75 to $12.00.
Cheapest paid models by input price
The eight lowest input prices start with IBM Granite 4.0 Micro at $0.017 input and $0.112 output per million tokens, 131,000 token context, open weights. Mistral Nemo follows at $0.019 input and $0.03 output, 131,072 tokens, open weights. inclusionAI Ling 3.0 Flash at $0.021 input and $0.063 output, 262,144 tokens, open weights. OpenAI GPT-5 Nano batch at $0.025 input and $0.20 output, 400,000 tokens, closed weights. Meta Llama 3.2 1B Instruct at $0.027 input and $0.201 output, 60,000 tokens, open weights. Upstage Solar Pro 4 at $0.03 input and $0.12 output, 524,288 tokens, closed weights. Qwen Qwen3.7 Flash at $0.03 input and $0.13 output, 1,000,000 tokens, closed weights. OpenAI gpt-oss-20b at $0.03 input and $0.13 output, 131,072 tokens, open weights.
Largest context windows
Five models share the 2,000,000 token maximum: OpenRouter Auto Router Beta, OpenRouter Pareto Code Router, SpaceXAI Grok 4.20 Multi-Agent, SpaceXAI Grok 4.20, and OpenRouter Auto Router.
Key takeaways
- 21 free models are available across seven providers with context windows up to 1,048,576 tokens.
- One model was removed and none were added or changed free status since yesterday.
- Ten models changed prices: four decreased input cost, four increased, two mixed.
- Cheapest paid input is IBM Granite 4.0 Micro at $0.017 per million tokens with open weights.
- Maximum context window across the catalog is 2,000,000 tokens on five router and Grok models.