OpenRouter API Pricing Update: 21 Free Models, DeepSeek Cuts Input Prices, New Entrants

The free tier holds 21 models, led by Google's Lyria 3 Pro Preview and Lyria 3 Clip Preview at 1,048,576 context tokens each. Thinking Machines offers Inkling Small and Inkling at the same context length. NVIDIA contributes five free models: Nemotron 3.5 Lightning, Nemotron 3 Ultra, Nemotron 3 Super, Nemotron 3 Nano Omni, and Nemotron 3.5 Content Safety, with context windows ranging from 128,000 to 1,000,000 tokens. Google adds Gemma 4 26B A4B and Gemma 4 31B at 262,144 tokens. Poolside provides Laguna S 2.1 and Laguna XS 2.1. inclusionAI offers Ling 3.0 Flash Sante and Ling 3.0 Flash Fin. Qwen contributes Qwen3.8 27B. Cohere lists North Mini Code at 256,000 tokens. Dots Studio has Dots3-Note Preview at 512,000 tokens. Space Bunny Alpha from stealth provider offers 1,000,000 tokens. OpenRouter's own Free Models Router provides 200,000 tokens.

Two new models appeared since 2026-09-25. Perceptron Mk1.5 from Perceptron enters at $0.15 per million input tokens and $1.50 per million output tokens with 36,864 context tokens. TypeSafe's Jev Router launches with no published pricing and a 1,000,000 token context window. Three models were removed: Nex AGI's Nex-N2.5-Mini and Nex-N2.5-Pro, both previously free, and Z.ai's GLM 5.2, also previously free.

DeepSeek: DeepSeek Pro Latest cut input from $0.3894 to $0.25 per million tokens while output rose from $1.1682 to $3.50. DeepSeek: DeepSeek Flash Latest reduced input from $0.04 to $0.035 and output from $1.00 to $0.29. DeepSeek: DeepSeek V4 Flash Latest dropped input from $0.03 to $0.022 with output steady at $0.32. DeepSeek: DeepSeek V4 Pro 0813 lowered input from $0.462 to $0.264 and output from $1.386 to $0.792. DeepSeek: DeepSeek V4 Flash 0423 trimmed input from $0.049 to $0.047 and output from $0.098 to $0.0941. Z.ai: GLM 5.1 made minor adjustments, input from $0.966 to $0.9646 and output from $3.036 to $3.0316. Z.ai: GLM 5.3 Flash held input at $0.045 while output fell from $0.60 to $0.14. Z.ai: GLM 4.7 increased input from $0.40 to $0.60 and output from $1.75 to $2.20. inclusionAI: Ling 3.0 Flash VL cut input from $0.06 to $0.021 and output from $0.18 to $0.0616. Google: Gemma 4 26B A4B reduced input from $0.09 to $0.0675 and output from $0.30 to $0.225. NVIDIA: Nemotron 3.5 Lightning lowered input from $0.08 to $0.07 with output unchanged at $0.20. Qwen: Qwen3 VL 30B A3B Instruct raised input from $0.13 to $0.15 and output from $0.52 to $0.60.

IBM: Granite 4.0 Micro leads at $0.017 per million input tokens and $0.112 output with 131,000 context tokens and open weights. OpenAI: gpt-oss-20b follows at $0.018 input and $0.09 output, 131,072 tokens, open weights. Mistral: Mistral Nemo at $0.019 input and $0.03 output, 131,072 tokens, open weights. inclusionAI: Ling 3.0 Flash VL at $0.021 input and $0.0616 output, 262,144 tokens, open weights. inclusionAI: Ling 3.0 Flash at $0.021 input and $0.063 output, same context, open weights. DeepSeek: DeepSeek V4 Flash Latest at $0.022 input and $0.32 output, 1,310,720 tokens, closed weights. DeepSeek: DeepSeek V4 Flash 0731 matches at $0.022 input and $0.32 output, same context, open weights. OpenAI: gpt-oss-20b batch variant at $0.024 input and $0.112 output, 131,072 tokens, open weights.

Five models share the maximum 2,000,000 token context window: OpenRouter's Auto Router (Beta), Pareto Code Router, and Auto Router; SpaceXAI's Grok 4.20 Multi-Agent and Grok 4.20.