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

Qwen-AgentWorld-35B-A3B

Qwen · moe · 34.7B parameters · 262,144 context

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

Parameters

34.7B

Context Window

256K tokens

Architecture

MoE

Best GPU

B200 SXM

Intelligence Brief

Qwen-AgentWorld-35B-A3B is a 34.7B parameter Mixture-of-Experts (256 experts, 8 active) model from Qwen, featuring Grouped Query Attention (GQA) with 40 layers and 2,048 hidden dimensions. With a 262,144 token context window, it supports tools, vision. For self-hosted inference, B200 SXM delivers optimal throughput at $4261/month.

Recent changes

Loading…

Related models

4 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

TypeMOE
Total Parameters34.7B
Active Parameters7B
Layers40
Hidden Dimension2,048
Attention Heads16
KV Heads2
Head Dimension256
Vocab Size248,320
Total Experts256
Active Experts8

Memory Requirements

BF16 Weights

69.4 GB

FP8 Weights

34.7 GB

INT4 Weights

17.4 GB

KV-Cache per Token81920 bytes
Activation Estimate0.00 GB

Fits on (single GPU) — most practical first

BEST FITA10G24GBINT4A3024GBINT4L424GBINT4RTX 409024GBINT4RTX 309024GBINT4RTX A500024GBINT4RX 7900 XTX24GBINT4RTX 509032GBINT4V100 32GB32GBINT4Instinct MI10032GBINT4TPU v432GBINT4TPU v6e (Trillium)32GBINT4

GPU Compatibility Matrix

Qwen-AgentWorld-35B-A3B is compatible with 57% of GPU configurations across 41 GPUs at 3 precision levels.

BF16 (Full)
FP8 (Half)
INT4 (Quarter)
Blackwell(7 GPUs)
B200 NVL (pair)360GB
81%
90%
95%
B300288GB
76%
88%
94%
B100 SXM192GB
64%
82%
91%
GB200 NVL72 (per GPU)192GB
64%
82%
91%
Hopper(7 GPUs)
H100 NVL 94GB (per GPU pair)188GB
63%
82%
91%
H200 SXM141GB
51%
75%
88%
H2096GB
28%
64%
82%
GH20096GB
28%
64%
82%
Ada Lovelace(11 GPUs)
L40S48GB
✗
28%
64%
L4048GB
✗
28%
64%
RTX 6000 Ada48GB
✗
28%
64%
L2048GB
✗
28%
64%
Ampere(16 GPUs)
A100 80GB SXM80GB
13%
57%
78%
A100 80GB PCIe80GB
13%
57%
78%
A1664GB
✗
46%
73%
RTX A600048GB
✗
28%
64%
Legend:No fitVery tightTightModerateGoodExcellent

GPU Recommendations

B200 SXMoptimal

FP8 · 1 GPU · tensorrt-llm

100/100

score

Throughput

1.1K tok/s

Latency (ITL)

1.0ms

Est. TTFT

0ms

Cost/Month

$4261

Cost/M Tokens

$1.54

Use this config →
H200 SXMoptimal

FP8 · 1 GPU · tensorrt-llm

100/100

score

Throughput

1.1K tok/s

Latency (ITL)

1.0ms

Est. TTFT

0ms

Cost/Month

$2553

Cost/M Tokens

$0.93

Use this config →
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 →

Deployment Options

API

API Deployment

No API pricing available

Self-Hosted

Single GPU

B200 SXM

$4261/mo

Min VRAM: 35 GB

Scale

Multi-GPU

RTX A6000 x2

411.6 tok/s

TP· $930/mo

Open Full Calculator →

API Pricing Comparison

No API pricing data available for this model.

Performance Estimates

Throughput by GPU

B200 SXM
1.1K tok/s
H200 SXM
1.1K tok/s
H100 SXM
1.1K tok/s

VRAM Breakdown (B200 SXM, FP8)

Weights
Weights 34.7 GBKV-Cache 0.7 GBActivations 0.0 GBOverhead 1.7 GB

Precision Impact

bf16

69.4 GB

weights/GPU

fp8

34.7 GB

weights/GPU

~1.1K tok/s

int4

17.4 GB

weights/GPU

Capabilities

Features

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

Supported Frameworks

vllm

Supported Precisions

BF16 (default)FP8INT4

Where to Deploy Qwen-AgentWorld-35B-A3B

Similar Models

Aya 23 35B

35B params · dense

Quality: 50

from $1.50/M

Similar specsCompare →

Command R (August 2024)

35B params · dense

Quality: 68

from $0.60/M

Smaller context, Higher qualityCompare →

Command R

35B params · dense

Quality: 68

from $0.50/M

Higher qualityCompare →

Yi 1.5 34B

34.4B params · dense

Quality: 72

from $0.80/M

Higher qualityCompare →

Code Llama 34B

34B params · dense

Quality: 55

from $0.78/M

Smaller contextCompare →

Frequently Asked Questions

How much VRAM does Qwen-AgentWorld-35B-A3B need for inference?

Qwen-AgentWorld-35B-A3B requires approximately 69.4 GB of VRAM at BF16 precision, 34.7 GB at FP8, or 17.4 GB at INT4 quantization. Additional VRAM is needed for KV-cache (81920 bytes per token) and activations (~0.00 GB).

What is the best GPU for Qwen-AgentWorld-35B-A3B?

The top recommended GPU for Qwen-AgentWorld-35B-A3B is the B200 SXM using FP8 precision. It achieves approximately 1.1K tokens/sec at an estimated cost of $4261/month ($1.54/M tokens). Score: 100/100.

How much does Qwen-AgentWorld-35B-A3B inference cost?

Qwen-AgentWorld-35B-A3B inference costs vary by provider and GPU setup. Use our calculator for detailed cost estimates across all providers.