Skip to content

Step 3.5 Flash on 4× H200 NVL: VRAM, speed and fit

Step 3.5 Flash at 8-bit runs on the 4× H200 NVL with 256K of context at about 176 tokens a second.

Text generation · Chat · Ollama / LM Studio

Runs comfortably · on 4× H200 NVL · 8-bit (Q8_0) · Ollama / LM Studio · Chat

GPU memory
297 GB
of 564 GB
Context
256K
auto
Speed
176 tok/s
27 when full
First reply
0.1 s
2K prompt

Where the memory goes

Model
197 GB
Context
96.6 GB
Overhead
2.7 GB
OS
0.0 GB
Free
267 GB
Memory
297 GB of 564 GB
Context
256K tokens · auto
Speed
176 tok/s · 27 when full
First reply
0.1 s for a 2K prompt
Running cost
$33/mo at home · $574/mo on a cloud 8× A40

Validation: Validated

What Step 3.5 Flash on 4× H200 NVL was checked against
CheckTargetResultStatus
Fits in GPU memory564 GB usable297 GB neededPass
Headroom5% free267 GB free (47%)Pass
Context for chat16K or more256KPass
Writing speed15 tok/s or more176 tok/sPass
First replyunder 3 s, 2K prompt0.1 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
8-bit (Q8_0)
Context
256K, auto
Chats at once
1
Use
Chat
Runs in
Ollama / LM Studio
Hardware
4× H200 NVL
GPU can use
564 GB of 564 GB

Running cost

At home
$33/mo
1520 W at the wall
Renting instead
$574/mo
8× A40 at $4.72/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 4× H200 NVL run?

Other models that run on 4× H200 NVL

Questions

Can Step 3.5 Flash run on 4× H200 NVL?

Step 3.5 Flash at 8-bit runs on the 4× H200 NVL with 256K of context at about 176 tokens a second.

Report VR-0BC62OA, calculated Nov 22, 2026 from published specs, not benchmarked.