LOADING DATASHEET
GROUNDED IN REAL HUMAN CONVERSATIONSYOUR BENCH
Nemotron Super
◆ OPEN
by Nvidia
OPEN WEIGHTSLOCALAPIREASONING
RANKED #256 OF 278 ASSISTANTS · OVERALL #402 OF 6,579 · VIBE SCORE 5.1 · 27 VOICES
Nemotron middle tier for collaborative agents and high-volume reasoning workloads
HAPPY5.6
3 mentions
SAD9.4
3 mentions
VIBE TRAJECTORY
26 weeks · 30 voices
HONEYMOON FADING
Early praise is cooling off in recent weeks.
HOW TO USE IT
Open weights. You can use a hosted service, or download it and run it yourself, free.
RUN IT ON YOUR OWN MACHINE
Open weights: download Nemotron Super once and it's yours. No subscription, no rate limits, works offline.
120B total parameters (12B active MoE), a Q4 quantized file is about 65 GB to 72 GB
- BARE MINIMUM
Mac Studio M2 Ultra (128 GB Unified Memory)
128 GB UNIFIED RAM
~$3500 USEDQ4 GGUF quant, ~18 tok/s, 8k context across unified memory
- SWEET SPOT
Workstation with 3x NVIDIA GeForce RTX 3090 24GB
72 GB VRAM
~$2400 USEDQ4_K_M quant with full GPU offload, ~28 tok/s, 16k context
- FULL POWER
Workstation with 4x NVIDIA RTX 6000 Ada 48GB
192 GB VRAM
~$28000 NEWFP8 / NVFP4 high precision, ~50 tok/s, full 128k+ context
LEARN HOW TO SET IT UP
WILL MY BOX RUN NEMOTRON SUPER?
The three cards above are researched picks. Search the machine you actually have — we only say yes if it should reply at a usable speed, not just load the weights and crawl.
HARDWARE PICKS & PRICES ARE RESEARCHED FROM THE LIVE WEB AND REFRESHED AUTOMATICALLY. TREAT THEM AS BALLPARK, NOT GOSPEL.
WHAT CAN YOU ACTUALLY DO WITH IT?
Everyday jobs on each of those rigs: how long Nemotron Super takes, and how good the crowd says it is at that kind of work.
| THE JOB | BARE MINIMUMMac Studio M2 Ultra (128 GB Unified Memory) | SWEET SPOTWorkstation with 3x NVIDIA GeForce RTX 3090 24GB | FULL POWERWorkstation with 4x NVIDIA RTX 6000 Ada 48GB |
|---|---|---|---|
| Write an emaila paragraph or two, drafted from a one-line brief | ~28 S★★★★★ | ~18 S★★★★★ | ~10 S★★★★★ |
| Summarize a documenta long report boiled down to the points that matter | ~83 S★★★★★ | ~54 S★★★★★ | ~30 S★★★★★ |
| Build a websitea small landing page, markup and styles together | ~7.4 MIN★★★★★ | ~4.8 MIN★★★★★ | ~2.7 MIN★★★★★ |
| Build a backendan API with routes, storage and tests | ~23 MIN★★★★★ | ~15 MIN★★★★★ | ~8.3 MIN★★★★★ |
| Build a gamea playable browser game in one file | ~14 MIN★★★★★ | ~8.9 MIN★★★★★ | ~5 MIN★★★★★ |
STARS ARE THE COMMUNITY SCORE FOR THAT KIND OF WORK, DOCKED FOR HOW SQUASHED THE WEIGHTS GET AT EACH TIER. TIMES ARE COMPUTED FROM RESEARCHED SPEEDS. BALLPARK, NOT GOSPEL.
WHAT THE INTERNET SAYS
People are mostly frustrated with Nemotron Super right now.
22 POSTS · 5 COMMENTS · GITHUB · REDDIT · X
PEOPLE LOVE
- No fan mail yet. The crowd is still warming up.
PEOPLE COMPLAIN ABOUT
- Outages, errors, or limits get in the way (5 mentions)
TOPIC BY TOPIC
- Working when you need itA SORE SPOT
0 thumbs up · 5 thumbs down
LOOKING FOR AN ALTERNATIVE?
Other models the crowd has fully reviewed, starting with text models like this one.
Qwen3.8 27B
DeepSeek V4 Flash> SOURCES (12)
- REDDIT · 2025-09-21 · 468 UPVOTESThere is a very real possibility that Google, OpenAI, Anthropic, etc. will release their own super cheap versions of Grok-4-fast!
There is a very real possibility that Google, OpenAI, Anthropic, etc. will release their own super cheap versions of Grok-4-fast!
- REDDIT · 2025-03-18 · 125 UPVOTESNew Nvidia Llama Nemotron Reasoning Models
New Nvidia Llama Nemotron Reasoning Models
- REDDIT · 2026-07-09 · 117 UPVOTESNVIDIA Puzzle-75B-A9B NVFP4 at 132 t/s on 3×3090 — Why is this size category a desert otherwise?
NVIDIA Puzzle-75B-A9B NVFP4 at 132 t/s on 3×3090 — Why is this size category a desert otherwise?
- REDDIT · 2026-03-12 · 31 UPVOTESWhat a joke...
What a joke...
- X · 2026-08-23 · 9 UPVOTESYou’re going to start seeing more and more companies catering open weights models based on the hardware out there. Nemot
You’re going to start seeing more and more companies catering open weights models based on the hardware out there. Nemot
- X · 2026-08-23 · 8 UPVOTESYes but then your stuck only using that model, that’s the issue. 27B is it, 2 x DGX Sparks gets you the ability to run D
Yes but then your stuck only using that model, that’s the issue. 27B is it, 2 x DGX Sparks gets you the ability to run D
- REDDIT · 2025-04-08 · 6 UPVOTESAnyone gotten Nemotron 49B Running in Ollama?
Anyone gotten Nemotron 49B Running in Ollama?
- REDDIT · 2026-04-19 · 6 UPVOTESFree Coding Agent with NVIDIA Nemotron (Open Source)
Free Coding Agent with NVIDIA Nemotron (Open Source)
- REDDIT · 2026-03-16 · 5 UPVOTESUnsloth, please create a Instruction-Following Dataset.
Unsloth, please create a Instruction-Following Dataset.
- GITHUB · 2026-09-07 · 3 UPVOTESAI team: add Grok 4.6 via Bedrock BYOK, drop the retired GLM free rung
AI team: add Grok 4.6 via Bedrock BYOK, drop the retired GLM free rung
- GITHUB · 2026-09-07 · 2 UPVOTESGrok 4.6 via Bedrock BYOK as the last-resort rung
Grok 4.6 via Bedrock BYOK as the last-resort rung
- GITHUB · 2026-09-07 · 0 UPVOTESfix(vllm): load Super-Omni RADIO final LayerNorm
fix(vllm): load Super-Omni RADIO final LayerNorm