NVIDIA
No datanemotron-3-ultra
Nemotron 3 Ultra
550B mixture-of-experts flagship with a 1M token context window. Best general-purpose choice for long documents and complex reasoning.
- Latency
- —
- 24h
- —
- Context
- 1M
ToolsStreaming
smart_toy
Free preview · no card required
Model Gateway
One API. Every model. Free preview.
Catalogue
All models
Every model the gateway exposes, with live health from the request log.
8 models available
DeepSeek
No datadeepseek-v4-flash
DeepSeek V4 Flash
Fast coding and reasoning model with strong SWE-bench performance. Good default for code generation and refactoring.
- Latency
- —
- 24h
- —
- Context
- 262K
ToolsStreamingReasoning
Cohere
No datanorth-mini-code
North Mini Code
30B mixture-of-experts model tuned for code. Low latency with a 256K context window.
- Latency
- —
- 24h
- —
- Context
- 256K
ToolsStreaming
InclusionAI
No dataling-3.0-flash
Ling 3.0 Flash
Lightweight high-throughput model for classification, extraction, and short-form generation.
- Latency
- —
- 24h
- —
- Context
- 262K
Streaming
NVIDIA
No datanemotron-3-nano-30b
Nemotron 3 Nano 30B
Compact 30B mixture-of-experts model. Efficient choice for summarisation, routing, and agent scaffolding.
- Latency
- —
- 24h
- —
- Context
- 256K
Streaming
Thinking Machines
No datainkling
Inkling
Reasoning model from Thinking Machines for long-horizon tasks. 256K context.
- Latency
- —
- 24h
- —
- Context
- 256K
StreamingReasoning
Meta
No datamuse-glimmer-30b
Muse Glimmer 30B
Meta image-and-text model. NIM free tier.
- Latency
- —
- 24h
- —
- Context
- 128K
Streaming
NVIDIA
No datanemotron-3.5-lightning-30b
Nemotron 3.5 Lightning 30B
Fast NVIDIA MoE for high-throughput reasoning. 256K context.
- Latency
- —
- 24h
- —
- Context
- 256K
Streaming
Catalogue updated 18 Aug, 08:54. Health, latency, and request volume come straight from the gateway request log. One OpenAI-compatible endpoint, every model — grab a free key.