UNIT · NVIDIA · GPU
24 GB VRAMworkstationReviewed June 2026

NVIDIA RTX PRO 4000 Blackwell

No editorial image yet — generic vendor mark shown. Credentials in spec table below.

Single-slot 140W Blackwell workstation card with 24GB GDDR7. The low-power, compact entry to the RTX PRO Blackwell line — fits small workstations and dense multi-card builds for local inference.

Released 2025·672 GB/s memory bandwidth
RUNLOCALAI SCORE
See full leaderboard →
455/ 1000
CC-tier
Estimated
Throughput
234/ 500
VRAM-fit
170/ 200
Ecosystem
200/ 200
Efficiency
46/ 100

Sub-scores sum to 650 / 1000. Headline = 650 × 0.70 (Estimated-confidence discount) = 455. This is an algorithmic performance-tier score — distinct from, and often lower than, the editorial “Our verdict” below, which weighs value and real-world fit (especially for hardware we haven’t measured yet). How scoring works →

Extrapolated from 672 GB/s bandwidth — 80.6 tok/s estimated. No measured benchmarks yet.

Plain-English: Workable at 32B, comfortable at 14B and below — snappy enough for a coding agent; vision models supported.

7B chat
Comfortable
14B chat
Comfortable
32B chat~
Tight
70B chat
Doesn't fit
Coding agent
Comfortable
Vision (≤8B VLM)
Comfortable
Long context (32K)
Comfortable
Comfortable — fits with headroom
~Tight — works, no slack
Marginal — needs aggressive quant
Doesn't fit usefully

Verdicts extrapolated from catalog VRAM + bandwidth + ecosystem flags. Hover any chip for the rationale. Want measured numbers? Submit your own run with runlocalai-bench --submit.

BLK · VERDICT

Our verdict

OP · Eruo Fredoline|VERIFIED JUN 18, 2026
7.3/10

What it does well

The RTX PRO 4000 Blackwell is the efficiency pick of the workstation line: 24GB CUDA VRAM in a single-slot, 140W card. That combination is rare and valuable — it drops into compact or SFF workstations, and its low power + single-slot width make it ideal for dense multi-card inference servers where a 4090/5090 would be too hot and wide. 24GB runs 32B-class models at Q4 and most diffusion workloads comfortably, with full CUDA and ECC.

Where it struggles

At ~$1,500 it's far pricier than a used RTX 3090 (also 24GB) or a new RTX 5070 Ti/5080-class card, and its 140W power budget caps raw throughput below those higher-wattage parts — you're paying for efficiency, single-slot density, ECC, and pro drivers, not speed. For a single hobbyist inference box, cheaper 24GB options give more tokens/sec/dollar.

Bottom line

The card to buy when you need 24GB of CUDA in a single slot at low power — compact workstations and multi-GPU inference racks. For a standalone budget 24GB build, a used 3090 remains the value king.

BLK · OVERVIEW

Overview

Single-slot 140W Blackwell workstation card with 24GB GDDR7. The low-power, compact entry to the RTX PRO Blackwell line — fits small workstations and dense multi-card builds for local inference.

Retailers we'd check:Amazon

Search-fallback link — editorial hasn't yet curated a retailer URL for this card.

Some links above are affiliate links. We may earn a commission at no extra cost to you. How we make money.

BLK · SPECS

Specs

VRAM24 GB
Power draw (peak)140 W
Released2025
MSRP$1500
Backends
CUDA
Vulkan

Models that fit

Open-weight models small enough to run on NVIDIA RTX PRO 4000 Blackwell with usable context.

Compare alternatives

Hardware worth comparing

The closest alternatives by price, memory bandwidth, and form factor, plus a step up and down — so you can frame the buying decision against real options.

Frequently asked

What models can NVIDIA RTX PRO 4000 Blackwell run?

With 24GB VRAM, the NVIDIA RTX PRO 4000 Blackwell runs models up to ~32B in 4-bit, with room for context. See the model list below for tested combinations.

Does NVIDIA RTX PRO 4000 Blackwell support CUDA?

Yes — NVIDIA RTX PRO 4000 Blackwell is an NVIDIA card with full CUDA support, the most mature local-AI backend. llama.cpp, Ollama, vLLM, and ExLlamaV2 all run natively.

Where next?

Reviewed by RunLocalAI Editorial. See our editorial policy for how we research and verify hardware specifications.