Skip to content
Updated minutes ago
Sber

GigaChat 20B

Sberbank · dense · 20B parameters · 8,192 context

Quality
50.0

Parameters

20B

Context Window

8K tokens

Architecture

Dense

Best GPU

H100 SXM

Intelligence Brief

GigaChat 20B is a 20B parameter DENSE model from Sberbank, featuring Grouped Query Attention (GQA) with 44 layers and 6,144 hidden dimensions. With a 8,192 token context window, it supports code, multilingual. For self-hosted inference, H100 SXM delivers optimal throughput at $1794/month.

Architecture Details

TypeDENSE
Total Parameters20B
Active Parameters20B
Layers44
Hidden Dimension6,144
Attention Heads48
KV Heads8
Head Dimension128
Vocab Size128,000

Memory Requirements

BF16 Weights

40.0 GB

FP8 Weights

20.0 GB

INT4 Weights

10.0 GB

KV-Cache per Token180224 bytes
Activation Estimate1.20 GB

GPU Compatibility Matrix

GigaChat 20B is compatible with 74% 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

H100 SXMoptimal

FP8 · 1 GPU · tensorrt-llm

100/100

score

Throughput

1.1K tok/s

Latency (ITL)

1.0ms

Est. TTFT

0ms

Cost/Month

$1794

Cost/M Tokens

$0.65

Use this config →
H20optimal

FP8 · 1 GPU · tensorrt-llm

100/100

score

Throughput

1.1K tok/s

Latency (ITL)

1.0ms

Est. TTFT

0ms

Cost/Month

$940

Cost/M Tokens

$0.34

Use this config →
H100 PCIeoptimal

FP8 · 1 GPU · tensorrt-llm

95/100

score

Throughput

836.6 tok/s

Latency (ITL)

1.2ms

Est. TTFT

0ms

Cost/Month

$1794

Cost/M Tokens

$0.82

Use this config →

Deployment Options

API

API Deployment

No API pricing available

Self-Hosted

Single GPU

H100 SXM

$1794/mo

Min VRAM: 20 GB

Scale

Multi-GPU

A100 40GB SXM x2

412.1 tok/s

TP· $1613/mo

API Pricing Comparison

No API pricing data available for this model.

Performance Estimates

Throughput by GPU

H100 SXM
1.1K tok/s
H20
1.1K tok/s
H100 PCIe
836.6 tok/s

VRAM Breakdown (H100 SXM, FP8)

Weights
Act
Weights 20.0 GBKV-Cache 1.5 GBActivations 9.6 GBOverhead 1.0 GB

Precision Impact

bf16

40.0 GB

weights/GPU

fp8

20.0 GB

weights/GPU

~1.1K tok/s

int4

10.0 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 GigaChat 20B

Similar Models

Larger context, Higher qualityCompare →

GPT-3.5 Turbo

20B params · dense

Quality: 67

from $1.50/M

Larger context, Higher qualityCompare →
Larger contextCompare →

Frequently Asked Questions

How much VRAM does GigaChat 20B need for inference?

GigaChat 20B requires approximately 40.0 GB of VRAM at BF16 precision, 20.0 GB at FP8, or 10.0 GB at INT4 quantization. Additional VRAM is needed for KV-cache (180224 bytes per token) and activations (~1.20 GB).

What is the best GPU for GigaChat 20B?

The top recommended GPU for GigaChat 20B is the H100 SXM using FP8 precision. It achieves approximately 1.1K tokens/sec at an estimated cost of $1794/month ($0.65/M tokens). Score: 100/100.

How much does GigaChat 20B inference cost?

GigaChat 20B inference costs vary by provider and GPU setup. Use our calculator for detailed cost estimates across all providers.