Compare / head-to-head

Qwen3-Next 80B-A3B ThinkingvsQwen3-VL 235B-A22B Instruct

Qwen3-Next 80B-A3B Thinking leads 7 of 9 shared benchmarks. Qwen3-Next 80B-A3B Thinking is 1.5x cheaper per token. Qwen3-Next 80B-A3B Thinking has the larger context window (262K tokens).

Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.

Shared benchmarks
7 – 2
Qwen3-Next 80B-A3B Thinking leads
Cheaper per token
Qwen3-Next 80B-A3B Thinking
1.5x cheaper, input + output
Larger context
Qwen3-Next 80B-A3B Thinking
262K tokens
Providers
Qwen
same provider
Qwen3-Next 80B-A3B Thinking
Qwen · released 2025-09-11
text
Context
262K
Max out
—
Input /1M
$0.15 ↗
Output /1M
$1.2 ↗
Cached /1M
—
Scores
16 · 6 core
Qwen3-VL 235B-A22B Instruct
Qwen · released 2025-09-23
textvisionvideo
Context
256K
Max out
—
Input /1M
$0.4 ↗
Output /1M
$1.6 ↗
Cached /1M
—
Scores
17 · 6 core
Quality

Benchmark matrix

BenchmarkQwen3-Next 80B-A3B ThinkingQwen3-VL 235B-A22B InstructΔEdge
Reported by both · 9
MMLU-Pro82.7% ↗81.8% ↗+0.9 ptQwen3-Next 80B-A3B Thinking
LMArena Elo1369.4 ↗1413.4 ↗-44Qwen3-VL 235B-A22B Instruct
AIME2587.8% ↗74.7% ↗+13.1 ptQwen3-Next 80B-A3B Thinking
Arena-Hard v262.3% ↗77.4% ↗-15.1 ptQwen3-VL 235B-A22B Instruct
BFCL-v372% ↗67.7% ↗+4.3 ptQwen3-Next 80B-A3B Thinking
HMMT2573.9% ↗57.4% ↗+16.5 ptQwen3-Next 80B-A3B Thinking
IFEval88.9% ↗87.8% ↗+1.1 ptQwen3-Next 80B-A3B Thinking
MMLU-Redux92.5% ↗92.2% ↗+0.3 ptQwen3-Next 80B-A3B Thinking
SuperGPQA60.8% ↗60.4% ↗+0.4 ptQwen3-Next 80B-A3B Thinking
Only Qwen3-Next 80B-A3B Thinking reports · 7
GPQA77.2% ↗not reported——
LiveBench 24112576.6% ↗not reported——
LiveCodeBench v6 (25.02-25.05)68.7% ↗not reported——
OJBench29.7% ↗not reported——
TAU2-Airline60.5% ↗not reported——
TAU2-Retail67.8% ↗not reported——
TAU2-Telecom43.9% ↗not reported——
Only Qwen3-VL 235B-A22B Instruct reports · 8
LCBV6 (25.02-25.05)not reported54.3% ↗——
LMArena Vision (LMArena)not reported1215.1 ↗——
MathVista (mini)not reported84.9% ↗——
MMMU (val)not reported78.7% ↗——
MMMU_Pronot reported68.1% ↗——
ScreenSpot Pronot reported62% ↗——
SimpleQAnot reported51.9% ↗——
VideoMME (w/o sub)not reported79.2% ↗——
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is Qwen3-Next 80B-A3B Thinking minus Qwen3-VL 235B-A22B Instruct in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing

Side by side

SpecQwen3-Next 80B-A3B ThinkingQwen3-VL 235B-A22B InstructEdge
Context window262K tokens256K tokensQwen3-Next 80B-A3B Thinking
Max output———
Input price / 1M$0.15 ↗$0.4 ↗Qwen3-Next 80B-A3B Thinking
Output price / 1M$1.2 ↗$1.6 ↗Qwen3-Next 80B-A3B Thinking
Cached input / 1M———
Input + output / 1M
Lower is cheaper. List prices; batch, tool and regional fees excluded.
$1.35$2Qwen3-Next 80B-A3B Thinking
Modalitiestexttext · vision · videoQwen3-VL 235B-A22B Instruct
Released2025-09-112025-09-23—
Cited benchmark scores1617—
Reliability

Qwen status

All providers →
Built by Respan
Which one wins on your data?

Public benchmarks are a starting point. Run Qwen3-Next 80B-A3B Thinking and Qwen3-VL 235B-A22B Instruct on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.