Models
Every model works with the same key and the same code. To switch, change the model name. All models are open-source models, run for BaseURL by Fireworks AI.
| Model | Powered by | Context (tokens) | Good for |
|---|---|---|---|
gateway-micro | Nemotron Lightning 3.5 30B | 262,144 | Cheapest. High-volume simple tasks: classification, extraction, short replies. |
gateway-fast | DeepSeek V4.1 Flash | 1,048,576 | Fast everyday tasks: chat, summaries, drafting. |
gateway-coder | gpt-oss-120b | 131,072 | Coding and reasoning at low cost. A reasoning model: give it a generous max_tokens. |
gateway-vision | GLM-5.3 Flash | 1,040,000 | Images and documents. |
gateway-smart | Qwen 3.8 Max | 262,144 | Smart all-round. |
kimi-k3 | Kimi K3 | 1,040,000 | Top quality, agents and coding. |
muse-spark-preview | Muse Spark (preview) | 1,048,576 | The newest release, offered early. Behaviour, availability and price may change. |
Current prices per million tokens are on the models table and, in naira, on the Models and pricing page of your dashboard. GET /v1/models lists the models your key can use.
Choosing a model
- Start with
gateway-fastfor most chat and text tasks. - Use
gateway-microwhen you send many small, simple requests and cost matters most. - Use
gateway-coderorkimi-k3for code and multi-step agents. - Reasoning models think before they answer and that thinking counts as output tokens. Give them a generous
max_tokens, or the answer may be cut short.
Questions? Email [email protected] with your X-Request-Id.