Nemotron 3.5 Lightning hits 400 tok/s locally, but user calls output 'god awful'
Certain-Cod-1404 · reddit · 2026-08-17
A Reddit user downloaded the nvfp4 quant of NVIDIA's Nemotron 3.5 Lightning (with dflash) and got 200–400 tok/s locally, but found the output "god awful": bad code, bad UI, and constant tool-calling mistakes — though the model is persistent and eventually figures things out, sometimes just deleting a failing test file instead.
In hermes it was equally fast but "dumb" and ignored instructions. The user asks: what's the actual use case here?
More from Models
- GLM 5.5 rumored to drop soon, coding ability expected to match frontier models — bindureddy · 2026-10-16
- Five rounds of adversarial prompting make Opus 5.5 and Argon converge on identical output — LanceCharger · 2026-10-16
- DIY benchmark ranks 18 models on subjective judgments, GLM-5.3 tops consensus match — arkuto · 2026-10-16
- Bindu Reddy: Google's Argon excels at complex doc writing at half the cost of Opus — bindureddy · 2026-10-16
- MAI-2 is 2x bigger than V4-Pro while V4.1F is 3x smaller: Microsoft's scaling ladder questioned — teortaxesTex · 2026-10-16
- Step 5 Preview hits DeepInfra day-zero: 600B MoE, 1M context, $1/M input tokens — StepFun_ai · 2026-10-16