Back to all LLMs

Qwen3 30B A3B Thinking

Support for 119 languages enabling multilingual RAG without translation overhead. Thinking mode with <think> blocks shows reasoning process over retrieved documents with Apache 2.0 license. If you want to compare the best LLMs for your data, try Agentset.

Leaderboard Rank
#15
of 16
ELO Rating
1250
#15
Win Rate
23.8%
#14
Latency
12313ms
#10

Model Information

Provider
Alibaba/Qwen
License
Open Source
Input Price per 1M
$0.05
Output Price per 1M
$0.34
Context Window
33K
Release Date
2025-08-28
Model Name
qwen3-30b-a3b-thinking-2507
Total Evaluations
1350

Performance Record

Wins321 (23.8%)
Losses870 (64.4%)
Ties159 (11.8%)
Wins
Losses
Ties

LLMs Are Just One Piece of RAG

Agentset gives you a managed RAG pipeline with the top-ranked models and best practices baked in. No infrastructure to maintain, no LLM orchestration to manage.

Trusted by teams building production RAG applications

5M+
Documents
1,500+
Teams
99.9%
Uptime

Performance Overview

ELO ratings by dataset

Qwen3 30B A3B Thinking's ELO performance varies across different benchmark datasets, showing its strengths in specific domains.

Qwen3 30B A3B Thinking - ELO by Dataset

Detailed Metrics

Dataset breakdown

Performance metrics across different benchmark datasets, including accuracy and latency percentiles.

SciFact

ELO 134329.3% WR132W-215L-103T

Quality Metrics

Correctness
4.97
Faithfulness
4.97
Grounding
4.93
Relevance
5.00
Completeness
4.87
Overall
4.95

Latency Distribution

Mean
8384ms
Min
2185ms
Max
19414ms

MSMARCO

ELO 130527.6% WR124W-276L-50T

Quality Metrics

Correctness
4.90
Faithfulness
4.90
Grounding
4.87
Relevance
5.00
Completeness
4.80
Overall
4.89

Latency Distribution

Mean
12522ms
Min
1541ms
Max
49799ms

PG

ELO 110414.4% WR65W-379L-6T

Quality Metrics

Correctness
4.90
Faithfulness
4.87
Grounding
4.83
Relevance
4.90
Completeness
4.67
Overall
4.83

Latency Distribution

Mean
16030ms
Min
3483ms
Max
44237ms

Build RAG in Minutes, Not Months

Agentset gives you a complete RAG API with top-ranked LLMs and smart retrieval built in. Upload your data, call the API, and get grounded answers from day one.

import { Agentset } from "agentset";

const agentset = new Agentset();
const ns = agentset.namespace("ns_1234");

const results = await ns.search(
  "What is multi-head attention?"
);

for (const result of results) {
  console.log(result.text);
}

Compare Models

See how it stacks up

Compare Qwen3 30B A3B Thinking with other top llms to understand the differences in performance, accuracy, and latency.