Fireworks AI Inference vs Nemotron 3.5 Super VL

Fireworks AI Inference

Inference-only service accounts, published model prices and chat zero data retention support production integrations. Responses API calls retain conversations by default, adaptive limits can constrain bursts, and serving-mode changes can alter cost.

Nemotron 3.5 Super VL

NVIDIA's free endpoint takes OpenAI-format requests with typed tool calls, JSON Schema output and an OpenAPI 3.1 file, and the weights run anywhere under a permissive licence. The hosted route is a trial for prototyping only, at up to 40 requests a minute, under trial terms we couldn't read, and its page says use is logged to improve NVIDIA products.

Fireworks AI Inference

1. Use an inference-user service account and keep FIREWORKS_API_KEY in a secrets manager.
2. Use https://api.fireworks.ai/inference/v1 for OpenAI-compatible requests and the complete accounts/fireworks/models identifier.
3. Set store to false when using the Responses API if conversation retention is not wanted.
4. Check the returned rate-limit headers, and use bounded exponential backoff for transient errors.
5. Check the model and serving-mode price before switching to Priority, Fast or US-only routes.

Nemotron 3.5 Super VL

1. Set chat_template_kwargs to {"enable_thinking": false} for short tasks. Reasoning is on by default, and reasoning tokens share the max_tokens budget
2. Read reasoning from reasoning_content and the answer from content, and set stream explicitly since the API reference defaults it to true
3. Don't send confidential or personal data to the hosted trial. The model page asks callers not to and says use is logged
4. Back off and retry on throttling. Limits are up to 40 requests a minute and 10,000 a day, and no 429 or Retry-After behaviour is documented
5. For production, use a partner endpoint or serve the weights. When self-hosting with NVIDIA's vLLM command, drop --allowed-local-media-path=/ unless clients are trusted