The real cost to run Mistral Medium 3.5 128B NVFP4

nvidia · 83.8B parameters · 8.2K context · license restricted

What it costs to run Mistral Medium 3.5 128B NVFP4 — $0 markup

• On your own machine — $0. Runs free locally if you have about 97 GB of VRAM at Q4_K_M (point Spanvero at LM Studio, Ollama, or llama.cpp).
• On your own rented GPU — from $0.65/hr. 5× NVIDIA Tesla P40 24GB at the direct Vast.ai price (≈ $1.06 a session), $0 markup.
• Via your own API key — $0.77/1M tokens (blended input + output). No GPU to manage (rough estimate for this size).

Mistral Medium 3.5 128B NVFP4 — 83.8B params (nvidia).

When does owning hardware beat the API?

At $0.77/1M tokens, the cheapest API for Mistral Medium 3.5 128B NVFP4 stays cheaper than Apple MacBook Pro 16" M3 Max 128GB (used) at any daily volume — electricity alone costs more per token than the API does.

Local is never $0 — every figure here includes electricity and hardware amortization. Speeds come from published 7-8B Q4 benchmarks (your model, quant, and context will differ); break-even volumes are rounded to 2 significant figures.

Mistral Medium 3.5 128B NVFP4 VRAM & system requirements

Open the free Spanvero advisor → for the live, interactive math for your exact workload and hardware.

Quick answers about Mistral Medium 3.5 128B NVFP4

How much does it cost to run Mistral Medium 3.5 128B NVFP4?

Three ways: $0 on your own machine if you have about 97 GB of VRAM (at Q4_K_M); a rented GPU from $0.65/hr (5× NVIDIA Tesla P40 24GB at the direct Vast.ai price); or about $0.77 per 1M blended tokens via your own API key (a rough estimate for this size). Spanvero adds $0 markup on every path.

Can I run Mistral Medium 3.5 128B NVFP4 locally?

Yes — Mistral Medium 3.5 128B NVFP4 runs locally for $0 in compute if your GPU or Apple Silicon Mac has about 97 GB of VRAM/unified memory (at Q4_K_M with 8.2K context). Point LM Studio, Ollama, or llama.cpp at it; after the download it runs offline and free.

What GPU do I need to run Mistral Medium 3.5 128B NVFP4?

Any GPU or Apple Silicon Mac with about 97 GB of VRAM/unified memory handles Mistral Medium 3.5 128B NVFP4 at Q4_K_M. Of the consumer hardware Spanvero tracks, the smallest that fits is Apple MacBook Pro 16" M3 Max 128GB (used).

Related models

• Qwen3.6 35B A3B NVFP4 — 18.7B, nvidia
• Gemma 4 31B IT NVFP4 — 20.9B, nvidia
• NVIDIA Nemotron 3 Super 120B A12B NVFP4 — 67.2B, nvidia
• DeepSeek-R1 — 671B, DeepSeek
• Qwen3 Coder 480B A35B Instruct — 480.2B, Qwen

Browse: More NVIDIA models · All models · Compare

Spanvero · What's new · Prices as of 2026-07-13. We're an honest advisor — $0 markup, your own accounts, we never resell compute. Catalog specs auto-indexed from the Hugging Face Hub — parameter count is exact; download size and quantizations are estimates. © 2026 Cynosure LLC.