StepFun, China · Released September 20, 2026New
Step-5-Preview has 604.3B parameters (33B active per token). In BF16 its weights alone take 1119 GiB; quantized to 4-bit, about 337 GiB. Each 32K-token request adds 2.28 GiB of KV cache. Cheapest way to run it today: 8 × RTX A6000 (48 GB) at $2.57 per hour. It reads text, images and video, calls tools and lets you set its reasoning effort.
Size your own setup in the calculator →vLLM needs the extra setup described on the model card (model card (opens in a new tab))
Weights plus the KV cache of one request (BF16 cache), in GiB. Add 1 to 3 GiB for the inference engine itself.
| Weights format | Weights only | + 8K context | + 32K context | + 128K context | + 1M context |
|---|---|---|---|---|---|
| BF16 / FP16 | 1119 | 1120 | 1122 | 1128 | 1189 |
| FP8 / INT8 | 561 | 561 | 563 | 570 | 631 |
| 4-bit (GGUF Q4_K_M) | 337 | 338 | 339 | 346 | 407 |
Fewest GPUs that fit the weights plus one 8K-token request. Data-center GPUs use vLLM (90% of memory usable), the others llama.cpp. Several GPUs split the model with tensor or pipeline parallelism.
| GPU | BF16 / FP16 | 8-bit | 4-bit |
|---|---|---|---|
| GeForce RTX 4090 24 GB | 64 (several machines) | 32 (several machines) | 16 (several machines) |
| GeForce RTX 5090 32 GB | 48 (several machines) | 24 (several machines) | 16 (several machines) |
| RTX PRO 6000 Blackwell 96 GB | 16 (several machines) | 7 | 4 |
| Mac M5 Max 128 GB 96 GB usable | 16 (several machines) | 6 | 4 |
| Ryzen AI Max+ 395 128 GB 96 GB usable | 16 (several machines) | 6 | 4 |
| DGX Spark (GB10) 128 GB | 16 (several machines) | 5 | 3 |
| NVIDIA H100 80 GB | 24 (several servers) | 16 (several servers) | 6 |
| NVIDIA H200 141 GB | 16 (several servers) | 5 | 3 |
| NVIDIA B200 (HGX) 180 GB | 8 | 4 | 3 |
8-bit: FP8 with vLLM (INT8 on GPUs without FP8), Q8_0 with llama.cpp. 4-bit: AWQ / GPTQ with vLLM, Q4_K_M with llama.cpp, both counted at the Q4_K_M size.
For each GPU: the fewest cards that fit Step-5-Preview with one 8K-token request, the most faithful weight format at that count, and the cheapest on-demand price (2026-10-16).
| Setup | Per hour | Weights | Provider | Per month | Rent |
|---|---|---|---|---|---|
| 8 × RTX A6000 48 GB | $2.57 | Q4_K_M | Vast.ai | $1,874.64 | Rent 8 × RTX A6000 on Vast.ai (opens in a new tab) → |
| 8 × RTX 6000 Ada 48 GB | $4.50 | Q4_K_M | Vast.ai | $3,287.92 | Rent 8 × RTX 6000 Ada on Vast.ai (opens in a new tab) → |
| 4 × RTX PRO 6000 Blackwell 96 GB | $5.34 | Q4_K_M | Vast.ai | $3,895.28 | Rent 4 × RTX PRO 6000 Blackwell on Vast.ai (opens in a new tab) → |
| 6 × NVIDIA A100 80GB | $5.61 | 4-bit | Vast.ai | $4,095.30 | Rent 6 × NVIDIA A100 80GB on Vast.ai (opens in a new tab) → |
| 3 × AMD MI300X 192 GB | $7.17 | 4-bit | RunPod | $5,234.10 | Rent 3 × AMD MI300X on RunPod (opens in a new tab) → |
| 3 × NVIDIA H200 141 GB | $8.69 | 4-bit | Vast.ai | $6,340.05 | Rent 3 × NVIDIA H200 on Vast.ai (opens in a new tab) → |
| 2 × AMD MI355X / MI350X 288 GB | $10.98 | 4-bit | RunPod | $8,015.40 | Rent 2 × AMD MI355X / MI350X on RunPod (opens in a new tab) → |
| 5 × NVIDIA H100 NVL 94 GB | $13.84 | 4-bit | Vast.ai | $10,103.20 | Rent 5 × NVIDIA H100 NVL on Vast.ai (opens in a new tab) → |
On-demand prices from AWS, Google Cloud, Azure, RunPod, Vast.ai, Oracle Cloud, OVHcloud, Scaleway, Verda, Akash and UpCloud, checked every hour. Serving many users needs more KV cache, so more memory: size it in the calculator. Some links are affiliate links: StudioTV may earn a commission, at no extra cost to you.
AWS, Google Cloud and UpCloud prices: Spare Cores (opens in a new tab), CC BY-SA 4.0 (opens in a new tab), shared under the same license.
Reasoning. Adjustable reasoning effort. Reasoning tokens take context and KV cache like any other token: leave room for them.
Architecture: 92 layers · 23 full attention (4 KV × 192) · sparse attention, a summary per 8-token block · 69 sliding-window layers (512 tokens, 4 KV × 192). Its KV cache is 3.8× smaller than if every layer used full attention, because it keeps only a short window of tokens at full detail and compresses the rest of the sequence. Only 33B parameters are active per token, but all 604B must sit in memory: VRAM depends on the total, speed on the active part.
| Context per request | BF16 cache | FP8 cache |
|---|---|---|
| 8K tokens | 0.65 GiB | 0.33 GiB |
| 32K tokens | 2.28 GiB | 1.15 GiB |
| 128K tokens | 8.82 GiB | 4.45 GiB |
| 1M tokens | 69.8 GiB | 35.3 GiB |
In BF16 the weights alone take 1119 GiB (1202 GB). In FP8 that is 561 GiB, and about 337 GiB with 4-bit quantization (Q4_K_M). Each request then adds KV cache: 0.65 GiB at 8K tokens and 2.28 GiB at 32K (BF16 cache).
No. Even in 4-bit it needs 16 RTX 4090s, spread over several machines. In 4-bit it needs 6 H100 GPUs.
No. Even in 4-bit it needs 16 RTX 5090s, spread over several machines.
Yes. In 4-bit (Q4_K_M with llama.cpp) it needs a Mac with 512 GB of unified memory, of which macOS gives the GPU about 384 GB, for one request with an 8K-token context. In 8-bit it does not fit on a single Mac.
No. Even in 4-bit it needs 3 DGX Spark units or 4 Ryzen AI Max+ 395 machines (128 GB), linked over the network.
On 2026-10-16, the cheapest on-demand setup is 8 × RTX A6000 (48 GB) with Q4_K_M weights on Vast.ai, at $2.57 per hour (about $1,874.64 per month). Next: 8 × RTX 6000 Ada (48 GB) with Q4_K_M weights on Vast.ai, at $4.50 per hour. Sized for one 8K-token request; prices are checked every hour.
With one request and an 8K-token context: 24 in BF16, 16 in FP8 and 6 in 4-bit (vLLM, 90% of memory usable). Above 8 GPUs, that means several servers. Serving many users at once needs more memory for their KV caches: the calculator sizes that for you.
Yes. Step-5-Preview reads text, images and video. Images add tokens to the context, so they also add KV cache.
Yes, Step-5-Preview supports tool (function) calling: it can ask your application to run a function and use the result.
Adjustable reasoning effort. Reasoning tokens count toward the context and the KV cache.
For a model card or a README: the 4-bit size and the smallest GPU it fits on, linked to this page. Data as JSON: /api/models/step-5-preview.json.
StepFun, China · Released September 20, 2026[](https://studiotvai.com/vram-requirements/step-5-preview)Same family (large moe): MiniMax M2.7 · DeepSeek R1 · Kimi K2.6 · GLM-5.3 · GLM-5.3-Flash · DeepSeek V4 Flash · DeepSeek V4 Pro · DeepSeek V4.1 Flash · MiMo-V2.6 Pro · MiMo-V2.6 Flash · Kimi K3 · Mistral Large 4 · DeepSeek V4 Flash Vision Exp · Qwen3.8 2.4T-A95B · Ornith 1.5 397B · Atria Dawn Preview · Intern S2 397B · DeepSeek V3 · DeepSeek V3.1 · Kimi K2.7 Code · Kimi K2.5 · Kimi K2 · GLM-5.2 · GLM-5.1 · GLM-5 · MiniMax M2.5 · MiniMax M2 · MiMo-V2.5 · Ornith-1.0 397B · Qwen3 235B-A22B · Qwen3.5 397B-A17B · GLM-4.7 · GLM-4.5 · Ling-3.1-flash
Llama 3.1 8B · Mistral Small 3.2 24B · Qwen3 32B · Llama 3.3 70B · Llama 3.1 405B · Qwen3.5 4B · Qwen3.5 9B · Qwen3.8 27B · Qwen3.6 35B-A3B · Qwen3.5 122B-A10B · Qwen3.8-Flash-Next · LFM2.5 8B-A1B · LFM2 24B-A2B · Gemma 4 12B · Gemma 4 31B · Gemma 4 26B-A4B · gpt-oss 20B · gpt-oss 120B · Llama 4 Scout 17B-16E · Llama 4 Maverick 17B-128E · Kolibri-1 · Nemotron 3 Nano 30B-A3B · Nemotron 3 Super 120B-A12B · MiniMax M2.7 · DeepSeek R1 · Kimi K2.6 · GLM-5.3 · GLM-5.3-Flash · DeepSeek V4 Flash · DeepSeek V4 Pro · DeepSeek V4.1 Flash · MiMo-V2.6 Pro · MiMo-V2.6 Flash · Kimi K3 · Mistral Large 4 · Qwen3-Coder-Next 80B-A3B · Qwen3-Coder 30B-A3B · Qwen3-Coder 480B-A35B · DeepSeek V4 Flash Vision Exp · Qwen3.8 2.4T-A95B · Ornith 1.5 397B · Ornith 1.5 35B-A3B · Ornith 1.5 9B · Atria Dawn Preview · Intern S2 397B · Clef · Clef Flash · D1 3B · Humanizer · JEV 27B VL · Mellum2.1 12B-A2.5B · LightOnOCR-3 4B · Spark-X2.5 4B · Agnes-3.0-Qwen · Youtu-Parsing-Omni · Agnes-3.0-Flash · Mistral 7B v0.3 · Mistral Small 3.1 24B · Mistral Nemo · Ministral 8B · DeepSeek V3 · DeepSeek R1 Qwen3 8B · DeepSeek R1 Distill Qwen 32B · DeepSeek V3.1 · DeepSeek R1 Distill Qwen 14B · DeepSeek R1 Distill Qwen 7B · DeepSeek R1 Distill Llama 8B · Qwen3 4B · Qwen3 8B · Qwen2.5 7B · Qwen3 14B · Qwen2.5 3B · Qwen3.6 27B · Qwen3.5 27B · Qwen2.5-Coder 7B · Qwen2.5 32B · Qwen2.5-Coder 14B · Qwen3.5 35B-A3B · Qwen3 30B-A3B · Qwen2.5 14B · Qwen2.5-Coder 32B · Kimi K2.7 Code · Kimi K2.5 · Kimi K2 · GLM-5.2 · GLM-5.1 · GLM-5 · MiniMax M2.5 · MiniMax M2 · MiMo-V2.5 · Ornith-1.0 397B · Qwen2.5-Coder 3B · Qwen3 235B-A22B · Qwen3.5 397B-A17B · Gemma 4 E4B · Gemma 4 E2B · Nemotron 3 Nano 4B · Nemotron 3.5 Lightning 30B-A3B · MiniCPM-V-4.7 35B-A3B · LFM2.5 24B-A2B · GLM-4.7 · Qwen3-VL 8B · Qwen2.5-VL 7B · Qwen3-VL 4B · Qwen2.5-VL 3B · Qwen2.5-VL 32B · Qwen2-VL 7B · GLM-4.5-Air · GLM-4.5 · phi-4 · Phi-3.5-mini · Phi-4-mini · Phi-3-mini-128k · Unlimited-OCR · Ling-3.1-flash
Estimates, not guarantees: computed from the official config.json with the same engine as the LLM VRAM Calculator. Real usage depends on your engine version and settings.