Fireworks AI Provider Guide
Open-model inference tuned for low-latency production workloads
Overview
Fireworks AI hosts Kimi, GPT-OSS, Qwen, GLM and
other open models with aggressive throughput optimizations.
NeuroLink talks to the OpenAI-compatible endpoint at api.fireworks.ai.
Key Facts
- Protocol: OpenAI-compatible (
/inference/v1/chat/completions) - Default base URL:
https://api.fireworks.ai/inference/v1 - Default model:
accounts/fireworks/models/kimi-k3(vision-capable) - Streaming: Yes
- Tool calling: Yes (model-dependent)
Quick Start
1. Get an API Key
https://fireworks.ai/account/api-keys
2. Configure Environment
FIREWORKS_API_KEY=fw_your-key
FIREWORKS_MODEL=accounts/fireworks/models/kimi-k3
3. Generate
import { NeuroLink } from "@juspay/neurolink";
const ai = new NeuroLink();
const result = await ai.generate({
provider: "fireworks",
input: { text: "Summarise the Raft consensus algorithm." },
});
console.log(result.content);
Supported Models (sample)
| Model ID | Vision | Notes |
|---|---|---|
accounts/fireworks/models/kimi-k3 | Yes | Default — newest Moonshot flagship |
accounts/fireworks/models/gpt-oss-120b | No | Text fallback (fallbackModelName) |
accounts/fireworks/models/qwen3p8-max | Yes | Qwen flagship |
accounts/fireworks/models/glm-5p3 | No | Rejects image inputs |
Fireworks lists models on /models that are not deployed for serverless
use, so the catalog only keeps ids that answered a real chat call.
kimi-k2p6, the default until September 2026, now returns 404
not deployed and is retired.
Browse: https://fireworks.ai/models
CLI Usage
pnpm run cli generate "..." --provider fireworks
pnpm run cli generate "..." --provider fireworks --model accounts/fireworks/models/qwen3p8-max
Provider Aliases
| Alias | Example |
|---|---|
fireworks | --provider fireworks |
Configuration Reference
| Environment Variable | Required | Default |
|---|---|---|
FIREWORKS_API_KEY | Yes | — |
FIREWORKS_MODEL | No | accounts/fireworks/models/kimi-k3 |
FIREWORKS_BASE_URL | No | https://api.fireworks.ai/inference/v1 |
Troubleshooting
Model not found, inaccessible, and/or not deployed— your account has not deployed the requested model. Check https://fireworks.ai/models and either deploy it or pick a serverless one.