Skip to content
Updated minutes ago
Rinna

Nekomata 14B

Rinna · dense · 14B parameters · 4,096 context

Quality
50.0

Parameters

14B

Context Window

4K tokens

Architecture

Dense

Best GPU

A100 40GB SXM

Intelligence Brief

Nekomata 14B is a 14B parameter DENSE model from Rinna, featuring Grouped Query Attention (GQA) with 40 layers and 5,120 hidden dimensions. With a 4,096 token context window, it supports multilingual. For self-hosted inference, A100 40GB SXM delivers optimal throughput at $807/month.

Architecture Details

TypeDENSE
Total Parameters14B
Active Parameters14B
Layers40
Hidden Dimension5,120
Attention Heads40
KV Heads8
Head Dimension128
Vocab Size65,536

Memory Requirements

BF16 Weights

28.0 GB

FP8 Weights

14.0 GB

INT4 Weights

7.0 GB

KV-Cache per Token163840 bytes
Activation Estimate1.00 GB

GPU Compatibility Matrix

Nekomata 14B is compatible with 82% 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

A100 40GB SXMoptimal

BF16 · 1 GPU · vllm

95/100

score

Throughput

299.9 tok/s

Latency (ITL)

3.3ms

Est. TTFT

1ms

Cost/Month

$807

Cost/M Tokens

$1.02

Use this config →
RTX A6000optimal

BF16 · 1 GPU · vllm

95/100

score

Throughput

148.1 tok/s

Latency (ITL)

6.8ms

Est. TTFT

1ms

Cost/Month

$465

Cost/M Tokens

$1.19

Use this config →
A40optimal

BF16 · 1 GPU · vllm

95/100

score

Throughput

134.2 tok/s

Latency (ITL)

7.5ms

Est. TTFT

1ms

Cost/Month

$399

Cost/M Tokens

$1.13

Use this config →

Deployment Options

API

API Deployment

No API pricing available

Self-Hosted

Single GPU

A100 40GB SXM

$807/mo

Min VRAM: 14 GB

Scale

Multi-GPU

RTX 3090 x2

298.7 tok/s

TP· $361/mo

API Pricing Comparison

No API pricing data available for this model.

Performance Estimates

Throughput by GPU

A100 40GB SXM
299.9 tok/s
RTX A6000
148.1 tok/s
A40
134.2 tok/s

VRAM Breakdown (A100 40GB SXM, BF16)

Weights
Act
Weights 28.0 GBKV-Cache 2.7 GBActivations 8.0 GBOverhead 2.2 GB

Precision Impact

bf16

28.0 GB

weights/GPU

~299.9 tok/s

fp8

14.0 GB

weights/GPU

int4

7.0 GB

weights/GPU

Capabilities

Features

Tool Use Vision Code Math Reasoning Multilingual Structured Output

Supported Frameworks

vllmtgi

Supported Precisions

BF16 (default)FP8INT4

Where to Deploy Nekomata 14B

Similar Models

Larger context, Higher qualityCompare →

RWKV-6 14B

14.1B params · hybrid

Quality: 50

from $0.20/M

Larger contextCompare →

Phi-4

14.7B params · dense

Quality: 73

from $0.14/M

Larger context, Higher qualityCompare →
Larger contextCompare →

Frequently Asked Questions

How much VRAM does Nekomata 14B need for inference?

Nekomata 14B requires approximately 28.0 GB of VRAM at BF16 precision, 14.0 GB at FP8, or 7.0 GB at INT4 quantization. Additional VRAM is needed for KV-cache (163840 bytes per token) and activations (~1.00 GB).

What is the best GPU for Nekomata 14B?

The top recommended GPU for Nekomata 14B is the A100 40GB SXM using BF16 precision. It achieves approximately 299.9 tokens/sec at an estimated cost of $807/month ($1.02/M tokens). Score: 95/100.

How much does Nekomata 14B inference cost?

Nekomata 14B inference costs vary by provider and GPU setup. Use our calculator for detailed cost estimates across all providers.