Nvidia — pricing & performance
Models served
9
of 1,302 tracked models
Mean output speed
55
tokens per second, averaged over
8 measured endpoints
8 measured endpoints
Mean latency
1.2s
time to first token, averaged over
8 measured endpoints
8 measured endpoints
Cheapest blended
$0.000
per 1M tokens at a 3:1
input:output blend
input:output blend
Output speed on Nvidia
Output tokens per second for the 8 of 9 models on Nvidia with published throughput telemetry, the most recent figure OpenRouter publishes for each host. The full table below lists all 9, measured or not. Higher is better. We do not run these measurements ourselves.
All 8 of 8 models⇩
The Known Good
9 models on Nvidia
Full leaderboard →
| Model | Creator | Input $/M | Output $/M | Blended | Speed | TTFT | Context |
|---|---|---|---|---|---|---|---|
NVIDIA: Nemotron 3 Nano 30B A3B (free)
|
NVIDIA | $0.000 | $0.000 | $0.000 | 111 | 0.5s | 256k |
|
NVIDIA: Nemotron 3 Nano Omni (free)
|
NVIDIA | $0.000 | $0.000 | $0.000 | 44 | 0.4s | 256k |
|
NVIDIA: Nemotron 3 Super (free)
|
NVIDIA | $0.000 | $0.000 | $0.000 | — | — | 262k |
|
NVIDIA: Nemotron 3 Ultra (free)
|
NVIDIA | $0.000 | $0.000 | $0.000 | 41 | 1.7s | 1.00M |
|
NVIDIA: Nemotron 3.5 Content Safety (free)
|
NVIDIA | $0.000 | $0.000 | $0.000 | 64 | 0.3s | 128k |
|
NVIDIA: Nemotron 3.5 Lightning (free)
|
NVIDIA | $0.000 | $0.000 | $0.000 | 69 | 0.5s | 1.00M |
|
NVIDIA: Nemotron 3.5 Super VL (free)
|
NVIDIA | $0.000 | $0.000 | $0.000 | 71 | 1.7s | 1.00M |
|
NVIDIA: Nemotron Nano 12B 2 VL (free)
|
NVIDIA | $0.000 | $0.000 | $0.000 | 11 | 3.7s | 128k |
|
NVIDIA: Nemotron Nano 9B V2 (free)
|
NVIDIA | $0.000 | $0.000 | $0.000 | 31 | 1.0s | 128k |
The Known Good
Source. Prices, endpoint availability and throughput for Nvidia are ingested from OpenRouter;
model metadata is enriched from models.dev. A blank cell means the figure was not published for that endpoint —
it never means zero. Blended price is (3 × input + output) ÷ 4, the same 3:1 blend used everywhere on this
site; see methodology. Quality scores are not provider-specific:
a model scores the same wherever it is hosted, so quality lives on the
model pages.