Qwen: Qwen3 30B A3B Thinking 2507

qwen · Released Aug 28, 2025
DFO score #240 / 749
13.5 leaderboard score
Speed #95 / 355
145.4 tok/s
Input Price #360 / 749
$0.200 per 1M tokens
Output Price #524 / 749
$2.40 per 1M tokens
Context #537 / 749
81,920 tokens

Analysis Summary

Qwen3 30B A3B Thinking 2507 is a Qwen reasoning model with a clear strength in mathematics: its math benchmark is substantially stronger than its overall intelligence and coding results. It supports tool use and function calling, and its 82K context window accommodates substantial documents and task context.

For agency workflows, it can suit math-heavy analysis, structured tasks, and tool-assisted automation where cost matters. The benchmark profile is uneven, with weaker coding and agentic indicators than the math result, so it is not a first choice for autonomous software engineering or complex multi-step work. Text-only input also limits use on visual client assets.

At the listed low input price, it is worth testing for bounded analytical tasks. Keep a stronger model available for writing or coding that needs consistent broad reasoning.

Assessed September 23, 2026

Editorial notes

Qwen3 30B A3B Thinking 2507 combines strong math results with tool and function calling, an 82K context window, and low input pricing. Its overall reasoning and coding results are more limited than the math scores suggest.

The DFO score ranks models on the Artificial Analysis Intelligence Index and is refreshed as new models launch. The write-up and verdict are our own. Feedback?

DFO Verdict

Qwen3 30B A3B Thinking 2507 combines strong math results with tool and function calling, an 82K context window, and low input pricing. Its overall reasoning and coding results are more limited than the math scores suggest.

#240 of 749 overall Down 87 this week

Benchmarks

Intelligence Index 9.2 #192 of 454 · Artificial Analysis
Coding Index 7.7 #167 of 199 · Artificial Analysis
Agentic Index 10.7 #84 of 181 · Artificial Analysis
Creative Writing Not rated Arena rating
Instruction Following Not rated Arena rating
GPQA Diamond 70.7%
HLE 9.8%
MMLU Pro 80.5%
MATH 500 97.6%
AIME 90.7%
AIME 2025 56.3%
SciCode 33.3%
LiveCodeBench 70.7%
TerminalBench Hard 5.3%
τ²-Bench 28.1%
IFBench 50.7%
LCR 59%

Artificial Analysis data refreshed Nov 16, 2026. Arena ratings from the Nov 9, 2026 text leaderboard, with style control. Bars: magenta = reasoning, ink = coding and agents, cyan = instructions and long context.

Artificial Analysis no longer lists this model, so its indexes are converted from their older scale to stay comparable.

How Qwen: Qwen3 30B A3B Thinking 2507 compares

Qwen: Qwen3 30B A3B Thinking 2507 ranks #192 of 454 AI models we track for overall intelligence, #167 of 199 for coding, #84 of 181 for agentic tasks. Its 82K-token context window is larger than 28% of the models we list. At $0.20 per million input tokens it is cheaper than 52% of comparable models.

Position in the field
DFO score: ahead of 68% of models #239
Speed: faster than 73% of models #95
Price: cheaper than 52% of models #360
Context: larger than 28% of models #537
worst in fieldmedianbest in field
Price vs frontier peers · $ per 1M tokens
Qwen: Qwen3 30B A3B Thinking 2507 $0.20 in $2.40 out
Anthropic: Claude Fable 5.2 $8.00 in $40.00 out
Google: Gemini 4 Carbon $2.00 in $10.00 out
Google: Gemini 4 Argon $2.00 in $10.00 out

Dark bar = input · light bar = output, scaled to the priciest peer.

Context window vs peers · tokens
Qwen: Qwen3 30B A3B Thinking 2507 82K

1M tokens ≈ 8 full-length novels or ~2,500 pages of business documents in a single request.

Pricing

Token Type Cost per 1M tokens Cost per 1K tokens
Input $0.20 $0.000200
Output $2.40 $0.002400

What would Qwen: Qwen3 30B A3B Thinking 2507 cost your business?

Pick the job that looks most like yours, then fine-tune with the sliders. Estimates update live.

A website chatbot handling around 100 customer conversations a day, a few short messages each.

3,000
One request is one message, email, draft or automation call.
1,200 tokens

$3.10/mo Qwen: Qwen3 30B A3B Thinking 2507 $0.0010 per request
$63.36/mo Anthropic: Claude Fable 5.2 $0.02 per request
$15.84/mo Google: Gemini 4 Carbon · best value $0.0053 per request

Full calculator with 749 models → Price Calculator

DFO AI AUTOMATION

These numbers get smaller with the right architecture.

We route routine calls to cheap models and save Qwen: Qwen3 30B A3B Thinking 2507 for the hard ones. Most clients cut their estimate by 60-80%.

Talk to our team

About Qwen: Qwen3 30B A3B Thinking 2507

Qwen3-30B-A3B-Thinking-2507 is a 30B parameter Mixture-of-Experts reasoning model optimized for complex tasks requiring extended multi-step thinking. The model is designed specifically for “thinking mode,” where internal reasoning traces are separated..

Frequently asked questions about Qwen: Qwen3 30B A3B Thinking 2507

How much does Qwen: Qwen3 30B A3B Thinking 2507 cost?

Qwen: Qwen3 30B A3B Thinking 2507 costs $0.20 per million input tokens and $2.40 per million output tokens.

What is the context window of Qwen: Qwen3 30B A3B Thinking 2507?

Qwen: Qwen3 30B A3B Thinking 2507 has a context window of 81,920 tokens (82K).

Is Qwen: Qwen3 30B A3B Thinking 2507 good for coding?

On our coding benchmark index, Qwen: Qwen3 30B A3B Thinking 2507 ranks #167 of 199 models, placing it in the broader range of the field for code generation and debugging.

What can Qwen: Qwen3 30B A3B Thinking 2507 do?

Qwen: Qwen3 30B A3B Thinking 2507 supports tool use and function calling.

Who created Qwen: Qwen3 30B A3B Thinking 2507?

Qwen: Qwen3 30B A3B Thinking 2507 is developed by Qwen and was released on August 28, 2025.

© 2026 Design for Online Ltd. Registered in England and Wales No. 10328553. VAT Registered. Design for Online® and Forerunner® are registered trademarks.