Hitting any supported model for a structured output is as simple as requesting a "response_format"

Get started with the example in our docs! openrouter.ai/docs/structu...

Multi-model routing is a lot easier if you can constrain parts of the output.

Supported models include OpenAI, Gemini (including 2.0), Cohere, Llama 3, QwQ, Mixtral, and more.

Browse here, and filter by price or context length: openrouter.ai/models?fmt=c...

ai.pydantic.dev/models/#api_...

🍊Mistral 7B Instruct v0.3: 50% cheaper, now $0.03 (was $0.055)
🍊Mistral Nemo: 70% cheaper, now $0.04 (was $0.13)
🍏Gemma 2 9B: 50% cheaper, now $0.03 (was $0.06)
👼Hermes 3B: $0.03 (was $0.14)
👼Hermes 70B: $0.20 (was $0.40)
👼Hermes 405B: $0.90 (was $1.80)

See the whole catalogue: openrouter.ai/models

🔷Qwen2.5 Coder 32B Instruct: 10% cheaper, now $0.08 (was $0.09)
🔷Qwen2.5 72B Instruct: 40% cheaper, now $0.23 (was $0.35)
🔷Qwen 32B Coder: $0.09 (was $0.18)

See them all on the new Author page: openrouter.ai/amazon

Incredible work! love seeing the live progress