| Nemotron 3.5 Lightning | GPT-6 Astra | |
|---|---|---|
| Input, per 1M tokens | $0.08 | $10.00 |
| Output, per 1M tokens | $0.20 | $50.00 |
| Context window | 1M | 1.05M |
| Classification or routing (600 tokens in, 10 out), per 1,000 | $0.05 | $6.50 |
| Chat reply (2,000 in, 400 out), per 1,000 | $0.24 | $40.00 |
| Retrieval-augmented answer (8,000 in, 500 out), per 1,000 | $0.74 | $105 |
| Confidence signal | yes | no |
| Tool calling | yes | yes |
| BANKING77, correct of 400 | 255 | not measured |
Which one should carry your traffic?
If Nemotron 3.5 Lightning answers your requests as well as GPT-6 Astra, moving that traffic saves 99% on chat-shaped requests. Averages cannot tell you whether it does for your requests. Nemotron 3.5 Lightning exposes a confidence signal, so Gimbix can let it answer first and hand unsure cases to GPT-6 Astra.
Nemotron 3.5 Lightning profile · GPT-6 Astra profile · Measure it on your traffic