Updated minutes ago
ReleasedJanuary 1, 2026source not yet verified
Other

siglip-so400m-14-384-flash-attn2-navit

HuggingFaceM4 · dense · 0.9B parameters · 8,192 context

Quality
50.0
Deploy this model →Use via API →Calculate ROI →Compare with othersFine-tune this model

Parameters

0.9B

Context Window

8K tokens

Architecture

Dense

Best GPU

RTX 4060

Intelligence Brief

siglip-so400m-14-384-flash-attn2-navit is a 0.9B parameter DENSE model from HuggingFaceM4, featuring Multi-Head Attention (MHA) with 27 layers and 1,152 hidden dimensions. With a 8,192 token context window, it supports vision. For self-hosted inference, RTX 4060 delivers optimal throughput at $209/month.

Recent changes

Loading…

Related models

3 suggestions

Picks: same family first, then same vendor within ±2× params, then top tag-overlap matches. Price shown is the cheapest Output $/M across providers — the row's page shows the canonical anchor.

Architecture Details

TypeDENSE
Total Parameters0.9B
Active Parameters0.9B
Layers27
Hidden Dimension1,152
Attention Heads16
KV Heads16
Head Dimension72
Vocab Size32,000

Memory Requirements

BF16 Weights

1.8 GB

FP8 Weights

0.9 GB

INT4 Weights

0.5 GB

KV-Cache per Token124416 bytes
Activation Estimate0.00 GB

Fits on (single GPU) — most practical first

BEST FITRTX 40608GBBF16RTX 30708GBBF16RTX 308010GBBF16RTX 4070 Ti12GBBF16RTX 4070 Super12GBBF16RTX 306012GBBF16A400016GBBF16RTX 408016GBBF16T416GBBF16V100 16GB16GBBF16RTX 4060 Ti 16GB16GBBF16A216GBBF16

GPU Compatibility Matrix

siglip-so400m-14-384-flash-attn2-navit is compatible with 100% of GPU configurations across 41 GPUs at 3 precision levels.

BF16 (Full)
FP8 (Half)
INT4 (Quarter)
Blackwell(7 GPUs)
B200 NVL (pair)360GB
100%
100%
100%
B300288GB
99%
100%
100%
B100 SXM192GB
99%
100%
100%
GB200 NVL72 (per GPU)192GB
99%
100%
100%
Hopper(7 GPUs)
H100 NVL 94GB (per GPU pair)188GB
99%
100%
100%
H200 SXM141GB
99%
99%
100%
H2096GB
98%
99%
100%
GH20096GB
98%
99%
100%
Ada Lovelace(11 GPUs)
L40S48GB
96%
98%
99%
L4048GB
96%
98%
99%
RTX 6000 Ada48GB
96%
98%
99%
L2048GB
96%
98%
99%
Ampere(16 GPUs)
A100 80GB SXM80GB
98%
99%
99%
A100 80GB PCIe80GB
98%
99%
99%
A1664GB
97%
99%
99%
RTX A600048GB
96%
98%
99%
Legend:No fitVery tightTightModerateGoodExcellent

GPU Recommendations

RTX 4060optimal

BF16 · 1 GPU · vllm

90/100

score

Throughput

734.4 tok/s

Latency (ITL)

1.4ms

Est. TTFT

0ms

Cost/Month

$209

Cost/M Tokens

$0.11

Use this config →
RTX 3070optimal

BF16 · 1 GPU · vllm

90/100

score

Throughput

1.2K tok/s

Latency (ITL)

0.8ms

Est. TTFT

0ms

Cost/Month

$85

Cost/M Tokens

$0.03

Use this config →
B200 SXMgood

BF16 · 1 GPU · tensorrt-llm

78/100

score

Throughput

3.5K tok/s

Latency (ITL)

0.3ms

Est. TTFT

0ms

Cost/Month

$4261

Cost/M Tokens

$0.46

Use this config →

Deployment Options

API

API Deployment

No API pricing available

Self-Hosted

Single GPU

RTX 4060

$209/mo

Min VRAM: 1 GB

Scale

Multi-GPU

RTX 4060

734.4 tok/s

Best available config

Open Full Calculator →

API Pricing Comparison

No API pricing data available for this model.

Performance Estimates

Throughput by GPU

RTX 4060
734.4 tok/s
RTX 3070
1.2K tok/s
B200 SXM
3.5K tok/s

VRAM Breakdown (RTX 4060, BF16)

Weights
KV
Weights 1.8 GBKV-Cache 2.0 GBActivations 0.0 GBOverhead 0.1 GB

Capabilities

Features

✗ Tool Use✓ Vision✗ Code✗ Math✗ Reasoning✗ Multilingual✗ Structured Output

Supported Frameworks

vllm

Supported Precisions

BF16 (default)

Where to Deploy siglip-so400m-14-384-flash-attn2-navit

Similar Models

Gemma 3 1B

1B params · dense

Quality: 35

Larger context, Lower qualityCompare →

Llama Guard 3 1B

1B params · dense

Quality: 50

Larger contextCompare →

Canary 1B

1B params · dense

Quality: 50

from $0.04/M

Similar specsCompare →

Qwen3-ASR-0.6B-hf

0.8B params · dense

Quality: 50

Larger contextCompare →

CSM-1B

1B params · dense

Quality: 50

Similar specsCompare →

Frequently Asked Questions

How much VRAM does siglip-so400m-14-384-flash-attn2-navit need for inference?

siglip-so400m-14-384-flash-attn2-navit requires approximately 1.8 GB of VRAM at BF16 precision, 0.9 GB at FP8, or 0.5 GB at INT4 quantization. Additional VRAM is needed for KV-cache (124416 bytes per token) and activations (~0.00 GB).

What is the best GPU for siglip-so400m-14-384-flash-attn2-navit?

The top recommended GPU for siglip-so400m-14-384-flash-attn2-navit is the RTX 4060 using BF16 precision. It achieves approximately 734.4 tokens/sec at an estimated cost of $209/month ($0.11/M tokens). Score: 90/100.

How much does siglip-so400m-14-384-flash-attn2-navit inference cost?

siglip-so400m-14-384-flash-attn2-navit inference costs vary by provider and GPU setup. Use our calculator for detailed cost estimates across all providers.