Skip to content
Updated minutes ago
NVIDIA

Eagle 2 9B

NVIDIA · dense · 9B parameters · 8,192 context

Quality
65.0

Parameters

9B

Context Window

8K tokens

Architecture

Dense

Best GPU

RTX 5090

Intelligence Brief

Eagle 2 9B is a 9B parameter DENSE model from NVIDIA, featuring Grouped Query Attention (GQA) with 32 layers and 4,096 hidden dimensions. With a 8,192 token context window, it supports vision. For self-hosted inference, RTX 5090 delivers optimal throughput at $845/month.

Architecture Details

TypeDENSE
Total Parameters9B
Active Parameters9B
Layers32
Hidden Dimension4,096
Attention Heads32
KV Heads8
Head Dimension128
Vocab Size151,936

Memory Requirements

BF16 Weights

18.0 GB

FP8 Weights

9.0 GB

INT4 Weights

4.5 GB

KV-Cache per Token131072 bytes
Activation Estimate0.50 GB

GPU Compatibility Matrix

Eagle 2 9B is compatible with 90% 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

RTX 5090optimal

BF16 · 1 GPU · vllm

95/100

score

Throughput

537.6 tok/s

Latency (ITL)

1.9ms

Est. TTFT

0ms

Cost/Month

$845

Cost/M Tokens

$0.60

Use this config →
V100 32GBoptimal

BF16 · 1 GPU · vllm

95/100

score

Throughput

93.7 tok/s

Latency (ITL)

10.7ms

Est. TTFT

2ms

Cost/Month

$180

Cost/M Tokens

$0.73

Use this config →
Instinct MI100optimal

BF16 · 1 GPU · vllm

95/100

score

Throughput

113.3 tok/s

Latency (ITL)

8.8ms

Est. TTFT

2ms

Cost/Month

$380

Cost/M Tokens

$1.28

Use this config →

Deployment Options

API

API Deployment

No API pricing available

Self-Hosted

Single GPU

RTX 5090

$845/mo

Min VRAM: 9 GB

Scale

Multi-GPU

A4000 x2

213.3 tok/s

TP· $323/mo

API Pricing Comparison

No API pricing data available for this model.

Performance Estimates

Throughput by GPU

RTX 5090
537.6 tok/s
V100 32GB
93.7 tok/s
Instinct MI100
113.3 tok/s

VRAM Breakdown (RTX 5090, BF16)

Weights
Act
Weights 18.0 GBKV-Cache 2.1 GBActivations 4.0 GBOverhead 1.4 GB

Precision Impact

bf16

18.0 GB

weights/GPU

~537.6 tok/s

fp8

9.0 GB

weights/GPU

int4

4.5 GB

weights/GPU

Capabilities

Features

Tool Use Vision Code Math Reasoning Multilingual Structured Output

Supported Frameworks

vllmtensorrt-llm

Supported Precisions

BF16 (default)FP8INT4

Where to Deploy Eagle 2 9B

Similar Models

Yi 1.5 9B

8.83B params · dense

Quality: 62

from $0.20/M

Similar specsCompare →
Larger context, Lower qualityCompare →

Gemma 2 9B

9.2B params · dense

Quality: 68

from $0.10/M

Similar specsCompare →

GLM-4 9B

9.4B params · dense

Quality: 50

from $0.15/M

Larger context, Lower qualityCompare →

Frequently Asked Questions

How much VRAM does Eagle 2 9B need for inference?

Eagle 2 9B requires approximately 18.0 GB of VRAM at BF16 precision, 9.0 GB at FP8, or 4.5 GB at INT4 quantization. Additional VRAM is needed for KV-cache (131072 bytes per token) and activations (~0.50 GB).

What is the best GPU for Eagle 2 9B?

The top recommended GPU for Eagle 2 9B is the RTX 5090 using BF16 precision. It achieves approximately 537.6 tokens/sec at an estimated cost of $845/month ($0.60/M tokens). Score: 95/100.

How much does Eagle 2 9B inference cost?

Eagle 2 9B inference costs vary by provider and GPU setup. Use our calculator for detailed cost estimates across all providers.