OpenRouter API Pricing Update: September 20, 2026
On September 20, 2026, OpenRouter tracked 447 models with 25 free models and 187 open-weight models, while several paid models saw input and output price reductions.
Free Model Availability
As of September 20, 2026, OpenRouter lists 25 models available at no cost, including Google's Lyria 3 Pro Preview and Lyria 3 Clip Preview, each with 1,048,576 context tokens. DeepSeek's DeepSeek V4 Flash 0731 (free) and Thinking Machines' Inkling Small and Inkling (free) also offer 1,048,576 context tokens at zero price.
NVIDIA's Nemotron 3.5 Lightning (free) and Nemotron 3 Ultra (free) provide 1,000,000 context tokens without charge. Additional free models from inclusionAI, Nex AGI, Qwen, Poolside, Google, NVIDIA, and Cohere offer context windows ranging from 256,000 to 262,144 tokens.
New and Removed Models
One new model was added on September 20, 2026: Z.ai's GLM 5.3 FlashX, priced at $0.37 per million input tokens and $1.25 per million output tokens, with a context window of 1,048,576 tokens and closed weights. No models were removed from the platform compared to September 19, 2026.
No models became free or stopped being free on this date, indicating stability in the free tier offerings.
Price Adjustments
Multiple models experienced price reductions on September 20, 2026. DeepSeek Flash Latest lowered input pricing from $0.135 to $0.13 per million tokens and output from $0.54 to $0.52. MoonshotAI's Kimi K3 reduced input from $1.95 to $1.7 and output from $10.92 to $8.5 per million tokens.
DeepSeek V4 Flash Latest cut input from $0.0519 to $0.04 and output from $0.1558 to $0.08 per million tokens. NVIDIA's Nemotron 3 Ultra decreased input from $0.625 to $0.6 and output from $3.125 to $2.4. Nemotron 3.5 Lightning reduced input from $0.08 to $0.07 per million tokens, with output unchanged at $0.2.
Further reductions include DeepSeek V4 Flash 0423 (input: $0.0484 → $0.0378; output: $0.0969 → $0.0756), Z.ai: GLM 5.2 (input: $0.5544 → $0.6496; output: $1.7424 → $2.0416 — note: increase), Z.ai: GLM Latest (input: $0.8918 → $0.8442; output: $2.8028 → $2.6532), MoonshotAI: Kimi Latest (same changes as Kimi K3), DeepSeek V4 Pro 0423 (input: $1.6 → $0.4223; output: $3.2 → $0.8446), and DeepSeek V4 Flash 0731 (input: $0.06 → $0.04; output: $0.12 → $0.08).
Cheapest Paid Options
The most affordable paid models by input cost are IBM's Granite 4.0 Micro at $0.017 per million input tokens and $0.112 for output, with 131,000 context tokens and open weights. Mistral Nemo follows at $0.019 input and $0.03 output, also open weights with 131,072 tokens.
Other low-cost paid options include inclusionAI's Ling 3.0 Flash ($0.021 input, $0.063 output, 262,144 tokens, open weights), OpenAI's GPT-5 Nano (batch) at $0.025 input and $0.2 output (400,000 tokens, closed weights), and Meta's Llama 3.2 1B Instruct ($0.027 input, $0.201 output, 60,000 tokens, open weights).
Largest Context Windows
The models with the largest context windows available on OpenRouter as of September 20, 2026, are Auto Router (Beta), Pareto Code Router, SpaceXAI's Grok 4.20 Multi-Agent, SpaceXAI's Grok 4.20, and Auto Router — all offering 2,000,000 context tokens. These are provided by openrouter and x-ai providers.
Key takeaways
- OpenRouter hosted 447 models on September 20, 2026, with 25 available for free and 187 featuring open weights.
- Price cuts were widespread among paid models, notably from DeepSeek, MoonshotAI, and NVIDIA, with no new free models added or removed.
- The least expensive paid model by input cost is IBM's Granite 4.0 Micro at $0.017 per million tokens.
- Z.ai's GLM 5.3 FlashX is the sole new model introduced, priced at $0.37 input and $1.25 output per million tokens.
- Five models share the maximum context length of 2,000,000 tokens, all accessible via OpenRouter or x-ai providers.