modelgrep
Q

Qwen: Qwen3 Next 80B A3B Thinking

qwen/qwen3-next-80b-a3b-thinking

87th smartest of 180Cheaper than 85% of paidReasoningToolsJSON
Use via OpenRouter ↗
Intelligence
26.7
87th of 180
Design Elo
Speed
168
23rd fastest
Latency
352ms
first token
Input price
$0.098
46th cheapest
Context
262K
33K max out

How it compares

Smarter than52%
of all ranked models
Faster than93%
of all ranked models
Cheaper than85%
of all ranked models

Overview

Qwen3-Next-80B-A3B-Thinking is a reasoning-first chat model in the Qwen3-Next line that outputs structured “thinking” traces by default. It’s designed for hard multi-step problems; math proofs, code synthesis/debugging, logic, and agentic...

Benchmarks

independent · via OpenRouter
Artificial Analysis47th percentile
Intelligence Index
26.7
Coding Index
19.5
Agentic Index
23.6
GPQA Diamond
76%
Humanity's Last Exam
12%
SciCode
39%
Tau²-Bench (agentic)
42%

Providers & pricing (5)

ProviderIn $/MOut $/MUptime
Alibaba$0.098$0.780
Nebiusfp8$0.150$1.20
Google$0.150$1.20
Novitabf16$0.150$1.50
AtlasCloudfp8$0.150$1.50

Specifications

Context window262K
Max output33K
Knowledge cutoffSep 2025
Input modalitiestext
Output modalitiestext
Prompt caching
Cache read price
ModeratedNo

Qwen3 Next 80B A3B Thinking FAQ

How much does Qwen3 Next 80B A3B Thinking cost?

Qwen3 Next 80B A3B Thinking costs $0.098 per million input tokens and $0.780 per million output tokens via OpenRouter, making it 46th cheapest of 298 paid models.

How smart is Qwen3 Next 80B A3B Thinking?

Qwen3 Next 80B A3B Thinking scores 26.7 on the Artificial Analysis Intelligence Index, ranking 87th of 180 benchmarked models, with a GPQA Diamond score of 76%.

How fast is Qwen3 Next 80B A3B Thinking?

Qwen3 Next 80B A3B Thinking generates around 168 tokens per second with 352ms time-to-first-token (p50), the 23rd fastest tracked model.

What is Qwen3 Next 80B A3B Thinking's context window?

Qwen3 Next 80B A3B Thinking supports a 262K-token context window and can output up to 33K tokens. It accepts text input.

Compare head-to-head