Skip to content
Updated minutes ago
Mistral

Codestral Mamba 7B

Mistral AI · hybrid · 7.3B parameters · 262,144 context

Quality
50.0

Parameters

7.3B

Context Window

256K tokens

Architecture

Dense

Best GPU

A30

Cheapest API

$0.60/M

Intelligence Brief

Codestral Mamba 7B is a 7.3B parameter HYBRID model from Mistral AI, featuring Multi-Head Attention (MHA) with 64 layers and 4,096 hidden dimensions. With a 262,144 token context window, it supports code, math. The most cost-effective API deployment is via mistral at $0.60/M output tokens. For self-hosted inference, A30 delivers optimal throughput at $332/month.

Architecture Details

TypeHYBRID
Total Parameters7.3B
Active Parameters7.3B
Layers64
Hidden Dimension4,096
Attention Heads1
KV Heads1
Head Dimension4096
Vocab Size32,768

Memory Requirements

BF16 Weights

14.6 GB

FP8 Weights

7.3 GB

INT4 Weights

3.6 GB

KV-Cache per Token0 bytes
Activation Estimate1.00 GB

GPU Compatibility Matrix

Codestral Mamba 7B is compatible with 95% of GPU configurations across 41 GPUs at 3 precision levels.

BF16 (Full)
FP8 (Half)
INT4 (Quarter)
Blackwell(7 GPUs)
B200 NVL (pair)360GB
B300288GB
B100 SXM192GB
GB200 NVL72 (per GPU)192GB
Hopper(7 GPUs)
H100 NVL 94GB (per GPU pair)188GB
H200 SXM141GB
H2096GB
GH20096GB
Ada Lovelace(11 GPUs)
L40S48GB
L4048GB
RTX 6000 Ada48GB
L2048GB
Ampere(16 GPUs)
A100 80GB SXM80GB
A100 80GB PCIe80GB
A1664GB
RTX A600048GB
Legend:No fitVery tightTightModerateGoodExcellent

GPU Recommendations

A30optimal

BF16 · 1 GPU · vllm

100/100

score

Throughput

310.6 tok/s

Latency (ITL)

3.2ms

Est. TTFT

1ms

Cost/Month

$332

Cost/M Tokens

$0.41

Use this config →
RTX 4090optimal

BF16 · 1 GPU · vllm

100/100

score

Throughput

335.5 tok/s

Latency (ITL)

3.0ms

Est. TTFT

1ms

Cost/Month

$370

Cost/M Tokens

$0.42

Use this config →
RTX 3090optimal

BF16 · 1 GPU · vllm

100/100

score

Throughput

311.6 tok/s

Latency (ITL)

3.2ms

Est. TTFT

1ms

Cost/Month

$180

Cost/M Tokens

$0.22

Use this config →

Deployment Options

API

API Deployment

mistral

$0.60/M

output tokens

Self-Hosted

Single GPU

A30

$332/mo

Min VRAM: 7 GB

Scale

Multi-GPU

RTX 3080 x2

392.5 tok/s

TP· $266/mo

API Pricing Comparison

ProviderInput $/MOutput $/MBadges
mistral$0.20$0.60
Cheapest

Cost Analysis

ProviderInput $/MOutput $/M~Monthly Cost
mistralBest Value$0.20$0.60$4

Cost per 1,000 Requests

Short (500 tok)

$0.22

via mistral

Medium (2K tok)

$0.88

via mistral

Long (8K tok)

$2.80

via mistral

Performance Estimates

Throughput by GPU

A30
310.6 tok/s
RTX 4090
335.5 tok/s
RTX 3090
311.6 tok/s

VRAM Breakdown (A30, BF16)

Weights
KV
Act
Weights 14.6 GBKV-Cache 17.2 GBActivations 8.0 GBOverhead 1.2 GB

Precision Impact

bf16

14.6 GB

weights/GPU

~310.6 tok/s

fp8

7.3 GB

weights/GPU

int4

3.6 GB

weights/GPU

Capabilities

Features

Tool Use Vision Code Math Reasoning Multilingual Structured Output

Supported Frameworks

vllmsglang

Supported Precisions

BF16 (default)FP8INT4

Where to Deploy Codestral Mamba 7B

Similar Models

Mistral 7B

7.3B params · dense

Quality: 56

from $0.07/M

Smaller context, Higher quality, CheaperCompare →
Smaller context, CheaperCompare →
Smaller context, CheaperCompare →

Frequently Asked Questions

How much VRAM does Codestral Mamba 7B need for inference?

Codestral Mamba 7B requires approximately 14.6 GB of VRAM at BF16 precision, 7.3 GB at FP8, or 3.6 GB at INT4 quantization. Additional VRAM is needed for KV-cache (0 bytes per token) and activations (~1.00 GB).

What is the best GPU for Codestral Mamba 7B?

The top recommended GPU for Codestral Mamba 7B is the A30 using BF16 precision. It achieves approximately 310.6 tokens/sec at an estimated cost of $332/month ($0.41/M tokens). Score: 100/100.

How much does Codestral Mamba 7B inference cost?

Codestral Mamba 7B API inference starts from $0.20/M input tokens and $0.60/M output tokens. Self-hosted inference costs depend on your GPU configuration — use our ROI calculator for a detailed breakdown.