OpenRouter LLM API Pricing Update: October 4, 2026

22 models are available at zero cost, including Google’s Lyria 3 Pro Preview and Lyria 3 Clip Preview, each with 1,048,576 context tokens. Thinking Machines offers two free Inkling variants with the same context length. NVIDIA provides three free Nemotron 3 models, two with 1,000,000 context tokens and one with 262,144. Additional free models come from Poolside, inclusionAI, Qwen, Apodex, Cohere, and others, with context windows ranging from 256,000 to 1,048,576 tokens.

Twelve models had price adjustments since October 3. DeepSeek Pro Latest saw input rise from $0.132 to $0.176 and output jump from $0.396 to $4.20 per million tokens. DeepSeek Flash Latest dropped input from $0.015 to $0.003 but increased output from $0.677 to $2.40. Z.ai’s GLM 5.3 cut input from $1.40 to $0.22 and output from $4.40 to $3.39. MoonshotAI’s Kimi Latest reduced input from $0.99 to $0.399, while Kimi K3 fell from $2.70 to $0.499 input. Z.ai’s GLM Latest halved input to $0.06 but doubled output to $8.00.

The lowest input cost among paid models is DeepSeek Flash Latest at $0.003 per million tokens, followed by DeepSeek V4 Flash Latest and V4 Flash 0731 at $0.0152. IBM’s Granite 4.0 Micro offers the lowest combined cost at $0.017 input and $0.112 output. OpenAI’s gpt-oss-20b is priced at $0.018 input and $0.09 output. Mistral Nemo charges $0.019 input and $0.03 output, the lowest output price in the list. All cheapest paid models support at least 131,072 context tokens, with three DeepSeek variants offering 1,048,576.