Mamba-Codestral 7B VRAM Requirements

Mamba-Codestral 7B at a glance

Made by

Mistral AI, France

Released

July 16, 2024

Parameters

7.29B

Context window

256K tokens

License

Apache 2.0, commercial use allowed

On Hugging Face

630 likes, 19K downloads last month (opens in a new tab)

VRAM by precision and context length

Weights plus the KV cache of one request (BF16 cache) and the model's fixed recurrent state, in GiB. Add 1 to 3 GiB for the inference engine itself.

Mamba-Codestral 7B GGUF files

Exact size of each quantization in gabriellarson/Mamba-Codestral-7B-v0.1-GGUF (opens in a new tab), and the smallest GPU that runs it with llama.cpp, Ollama or LM Studio and an 8K-token context.

Read every day from Hugging Face. Split files are added together; vision projectors are left out.

How many GPUs to run Mamba-Codestral 7B

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.

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.

Cheapest way to run Mamba-Codestral 7B today

For each GPU: the fewest cards that fit Mamba-Codestral 7B with one 8K-token request, the most faithful weight format at that count, and the cheapest on-demand price (2026-10-30).

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.

KV cache: how memory grows with context

Architecture: 64 Mamba-2 layers, no attention: fixed state instead of a KV cache. It has no attention layers, so no KV cache: each request holds a fixed 260 MiB state, whatever the context.

Plus 260 MiB of recurrent state per request, whatever the context.

Mamba-Codestral 7B VRAM badge

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/mamba-codestral-7b.json.

[![Mamba-Codestral 7B VRAM](https://studiotvai.com/badge/mamba-codestral-7b.svg)](https://studiotvai.com/vram-requirements/mamba-codestral-7b)

Other models

Same family (hybrid mamba): Nemotron 3 Nano 30B-A3B · Nemotron 3 Super 120B-A12B · Nemotron 3 Ultra 550B-A55B · Nemotron 3.5 Super VL 120B-A12B · Nemotron 3.5 Bolt 74B-A9B · Nemotron 3 Nano 4B

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.