Inference provider

OpenRouter

A multi-provider inference gateway with configurable provider routing.

Models tracked28
Route-ready28
Input from$0.030
Output from$0.080

Routing contract

Modegateway
Protocolopenai
API basehttps://openrouter.ai/api/v1
Last observedAug 28, 2026
Routing groundwork only

These public route targets identify where and how a model can be requested. Model Markets does not hold provider keys or send inference traffic from this interface.

Provider records describe a serving surface, not the underlying model. Pricing, quantization, region, service tier, retention policy, and measured performance remain endpoint-specific.

Catalog coverage

Model endpoints

Provider website ↗
ModelRoute modelContextInput / 1MOutput / 1MRegionEvidence
gpt-oss-20bRoute-readyopenai/gpt-oss-20b128K$0.030$0.13Provider-reported
MiniMax-M2.7Route-readyminimax/minimax-m2.7200K$0.30$1.20Provider-reported
Llama-3.1-8B-InstructRoute-readymeta-llama/llama-3.1-8b-instruct128K$0.050$0.080Provider-reported
gpt-oss-120bRoute-readyopenai/gpt-oss-120b128K$0.037$0.17Provider-reported
Qwen3.8-27BRoute-readyqwen/qwen3.8-27b256K$0.425$2.55Provider-reported
DeepSeek-R1Route-readydeepseek/deepseek-r1160K$0.70$2.50Provider-reported
Kimi-K3Route-readymoonshotai/kimi-k31M$3.00$15.00Provider-reported
Qwen3.5-35B-A3BRoute-readyqwen/qwen3.5-35b-a3b256K$0.25$1.25Provider-reported
GLM-5.2Route-readyz-ai/glm-5.21M$1.19$3.74Provider-reported
DeepSeek-V4-FlashRoute-readydeepseek/deepseek-v4-flash1M$0.0868$0.1736Provider-reported
DeepSeek-V3.2Route-readydeepseek/deepseek-v3.2160K$0.269$0.40Provider-reported
DeepSeek-V4-ProRoute-readydeepseek/deepseek-v4-pro1M$0.772$1.5441Provider-reported
Kimi-K2.6Route-readymoonshotai/kimi-k2.6256K$0.95$4.00Provider-reported
phi-4Route-readymicrosoft/phi-416K$0.070$0.14Provider-reported
Kimi-K2.5Route-readymoonshotai/kimi-k2.5256K$0.60$3.00Provider-reported
Llama-3.1-70B-InstructRoute-readymeta-llama/llama-3.1-70b-instruct128K$0.40$0.40Provider-reported
MiniMax-M2.5Route-readyminimax/minimax-m2.5192K$0.27$1.08Provider-reported
Qwen3-Coder-NextRoute-readyqwen/qwen3-coder-next256K$0.12$0.80Provider-reported
Llama-3.3-70B-InstructRoute-readymeta-llama/llama-3.3-70b-instruct128K$0.71$0.71Provider-reported
Qwen3.5-397B-A17BRoute-readyqwen/qwen3.5-397b-a17b256K$0.39$2.34Provider-reported
MiniMax-M3Route-readyminimax/minimax-m31M$0.30$1.20Provider-reported
GLM-5.1Route-readyz-ai/glm-5.1198K$1.26$3.96Provider-reported
GLM-5Route-readyz-ai/glm-5198K$0.60$1.92Provider-reported
Kimi-K2-ThinkingRoute-readymoonshotai/kimi-k2-thinking256K$0.60$2.50Provider-reported
Qwen3.8-2.4T-A95BRoute-readyqwen/qwen3.8-2.4t-a95b256K$2.00$6.00Provider-reported
Hy3Route-readytencent/hy3256K$0.0825$0.33Provider-reported
granite-4.1-8bRoute-readyibm-granite/granite-4.1-8b128K$0.050$0.10Provider-reported
Olmo-3-32B-ThinkRoute-readyallenai/olmo-3-32b-think64K$0.15$0.50Provider-reported

Questions

OpenRouter questions

Which models does OpenRouter offer?+

The current source snapshot includes 28 model endpoint for OpenRouter: gpt-oss-20b, MiniMax-M2.7, Llama-3.1-8B-Instruct, gpt-oss-120b, Qwen3.8-27B, DeepSeek-R1, Kimi-K3, Qwen3.5-35B-A3B, GLM-5.2, DeepSeek-V4-Flash, DeepSeek-V3.2, DeepSeek-V4-Pro, Kimi-K2.6, phi-4, Kimi-K2.5, Llama-3.1-70B-Instruct, MiniMax-M2.5, Qwen3-Coder-Next, Llama-3.3-70B-Instruct, Qwen3.5-397B-A17B, MiniMax-M3, GLM-5.1, GLM-5, Kimi-K2-Thinking, Qwen3.8-2.4T-A95B, Hy3, granite-4.1-8b, Olmo-3-32B-Think. Browse all models

How much does OpenRouter cost?+

For gpt-oss-20b, the cited source reports $0.030 input and $0.13 output per million tokens.

Is OpenRouter the fastest provider?+

Model Markets has not run approved independent probes for this delivery, so it does not make a fastest-provider claim.

How current is OpenRouter data?+

This source snapshot was observed Aug 28, 2026 and is marked non-live until production collectors are activated.

How does Model Markets verify OpenRouter?+

Catalog and price claims retain their source classification. Independent telemetry will use disclosed-region standardized probes and will be presented separately. Read the methodology