Migrate from Fireworks AI to Railwail

TL;DR — Switch in Under 5 Minutes

• Both APIs are OpenAI-compatible — change base URL and key only
• All Fireworks-hosted models mirrored: Llama 3.3, Mixtral, DeepSeek, Qwen, FireFunction
• EU-hosted endpoint, EUR billing
• Same low-latency vLLM-based serving for open models
• Plus access to closed-source frontier models (Claude, GPT-4o, Gemini) on the same key

Why Move Off Fireworks AI?

Fireworks AI is well-known for fast inference on open-source LLMs. The trade-offs are US-only hosting, USD billing, and the lack of closed-source frontier models (Claude, GPT-4o, Gemini) in the catalog. Railwail keeps the speed and adds EU hosting, EUR billing, and the closed-source models — all behind one API key.

Step 1 — Get a Railwail API Key

Sign up at railwail.com and generate a key. Free credits included.

Step 2 — Change Base URL and Model Slug

TypeScript / JavaScript

Before (Fireworks):

import OpenAI from "openai";

const fireworks = new OpenAI({
  apiKey: process.env.FIREWORKS_API_KEY,
  baseURL: "https://api.fireworks.ai/inference/v1",
});

const res = await fireworks.chat.completions.create({
  model: "accounts/fireworks/models/llama-v3p3-70b-instruct",
  messages: [{ role: "user", content: "Hello" }],
});

import OpenAI from "openai";

const client = new OpenAI({
  apiKey: process.env.RAILWAIL_API_KEY,
  baseURL: "https://api.railwail.com/v1",
});

const res = await client.chat.completions.create({
  model: "llama-3.3-70b-instruct",
  messages: [{ role: "user", content: "Hello" }],
});

Python

from openai import OpenAI

client = OpenAI(
    api_key=os.environ["RAILWAIL_API_KEY"],
    base_url="https://api.railwail.com/v1",
)

resp = client.chat.completions.create(
    model="llama-3.3-70b-instruct",
    messages=[{"role": "user", "content": "Hello"}],
)
print(resp.choices[0].message.content)

cURL

curl https://api.railwail.com/v1/chat/completions \
  -H "Authorization: Bearer $RAILWAIL_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "llama-3.3-70b-instruct",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

API Endpoint Mapping

Model Mapping

Pricing Comparison (per 1M tokens, May 2026)

Why Railwail Over Fireworks

• EU billing in EUR with VAT receipts
• Frankfurt-region hosting for low-latency EU customers
• Same OpenAI-compatible API
• Adds Claude, GPT-4o, Gemini to the catalog — Fireworks has no closed-source
• Built-in playground for cross-model comparison
• Comparable pricing on Fireworks-style open models

FAQ

What about FireFunction and the function-calling models?

FireFunction v2 is mirrored as firefunction-v2 with identical tool/function calling behaviour. For higher-quality function calling, llama-3.3-70b-instruct now matches FireFunction v2 on most benchmarks.

Can I bring my own LoRA?

Custom LoRAs (Fireworks' multi-LoRA feature) are not currently hosted on Railwail. You can keep them on Fireworks and use Railwail for everything else.

What about Fireworks' speculative decoding speedups?

Railwail uses vLLM with speculative decoding on all Llama models — comparable throughput to Fireworks.

How does streaming compare?

Both APIs stream OpenAI-format SSE chunks. First-token latency from EU origins is typically lower via Railwail's Frankfurt edge.

What about embeddings?

POST /v1/embeddings accepts identical request shapes. Railwail adds Voyage and Cohere embedding models alongside the open-source options.

Next Steps

• Sign up at railwail.com
• Generate an API key
• Update baseURL to https://api.railwail.com/v1
• Replace long Fireworks model paths with short Railwail slugs
• Read the reference at railwail.com/docs
• Compare per-token pricing at railwail.com/pricing