Skip to content

Step 3.5 Flash on 2× RTX PRO 6000 Blackwell: VRAM, speed and fit

Step 3.5 Flash at 6-bit runs on the 2× RTX PRO 6000 Blackwell with 72K of context at about 113 tokens a second.

Text generation · Chat · Ollama / LM Studio

Runs comfortably · on 2× RTX PRO 6000 Blackwell · 6-bit (Q6_K) · Ollama / LM Studio · Chat

GPU memory
181 GB
of 192 GB
Context
72K
auto
Speed
113 tok/s
31 when full
First reply
0.3 s
2K prompt

Where the memory goes

Model
152 GB
Context
27.6 GB
Overhead
1.5 GB
OS
0.0 GB
Free
10.7 GB
Memory
181 GB of 192 GB
Context
72K tokens · auto
Speed
113 tok/s · 31 when full
First reply
0.3 s for a 2K prompt
Running cost
$18/mo at home · $287/mo on a cloud 4× A40

Validation: Validated

What Step 3.5 Flash on 2× RTX PRO 6000 Blackwell was checked against
CheckTargetResultStatus
Fits in GPU memory192 GB usable181 GB neededPass
Headroom5% free10.7 GB free (6%)Pass
Context for chat16K or more72KPass
Writing speed15 tok/s or more113 tok/sPass
First replyunder 3 s, 2K prompt0.3 sPass

Targets: a reply you can read along with, a first reply the use case tolerates, and 5% of memory left free. Calculated from published specs, not benchmarked.

The setup

Model
Step 3.5 Flash
Parameters
199B, 9.4B active
Precision
6-bit (Q6_K)
Context
72K, auto
Chats at once
1
Use
Chat
Runs in
Ollama / LM Studio
Hardware
2× RTX PRO 6000 Blackwell
GPU can use
192 GB of 192 GB

Running cost

At home
$18/mo
800 W at the wall
Renting instead
$287/mo
4× A40 at $2.36/h

4 h a day, power at $0.18 per kWh, cloud at RunPod on-demand prices as of Nov 22, 2026.

How to run it

llama.cpp
llama-server -hf ggml-org/Step-3.5-Flash-GGUF
vLLM
vllm serve stepfun-ai/Step-3.5-Flash
MLX (Mac)
mlx_lm.chat --model mlx-community/Step-3.5-Flash-4bit

Step 3.5 Flash on similar machines

All machines: Step 3.5 Flash VRAM requirements. Everything that runs here: what can the 2× RTX PRO 6000 Blackwell run?

Other models that run on 2× RTX PRO 6000 Blackwell

Questions

Can Step 3.5 Flash run on 2× RTX PRO 6000 Blackwell?

Step 3.5 Flash at 6-bit runs on the 2× RTX PRO 6000 Blackwell with 72K of context at about 113 tokens a second.

Report VR-1ALMD98, calculated Nov 22, 2026 from published specs, not benchmarked.