Compare / head-to-head
Qwen3 4B Thinking 2507vsQwen3-Next 80B-A3B Instruct
Qwen3-Next 80B-A3B Instruct leads 9 of 13 shared benchmarks. Both offer a 262K-token context window.
Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.
Shared benchmarks
4 – 9
Qwen3-Next 80B-A3B Instruct leads
Cheaper per token
—
list price, input + output
Larger context
Tie
both 262K tokens
Providers
Qwen
same provider
Qwen3 4B Thinking 2507
Qwen · released 2025-08-06
text
- Context
- 262K
- Max out
- —
- Input /1M
- —
- Output /1M
- —
- Cached /1M
- —
- Scores
- 15 · 5 core
Qwen3-Next 80B-A3B Instruct
Qwen · released 2025-09-11
text
Quality
Benchmark matrix
13 shared · 2 only Qwen3 4B Thinking 2507 · 3 only Qwen3-Next 80B-A3B Instruct| Benchmark | Qwen3 4B Thinking 2507 | Qwen3-Next 80B-A3B Instruct | Δ | Edge |
|---|---|---|---|---|
| Reported by both · 13 | ||||
| MMLU-Pro | 74% ↗ | 80.6% ↗ | -6.6 pt | Qwen3-Next 80B-A3B Instruct |
| AIME25 | 81.3% ↗ | 69.5% ↗ | +11.8 pt | Qwen3 4B Thinking 2507 |
| Arena-Hard v2 | 34.9% ↗ | 82.7% ↗ | -47.8 pt | Qwen3-Next 80B-A3B Instruct |
| BFCL-v3 | 71.2% ↗ | 70.3% ↗ | +0.9 pt | Qwen3 4B Thinking 2507 |
| GPQA | 65.8% ↗ | 72.9% ↗ | -7.1 pt | Qwen3-Next 80B-A3B Instruct |
| HMMT25 | 55.5% ↗ | 54.1% ↗ | +1.4 pt | Qwen3 4B Thinking 2507 |
| IFEval | 87.4% ↗ | 87.6% ↗ | -0.2 pt | Qwen3-Next 80B-A3B Instruct |
| LiveBench 20241125 | 71.8% ↗ | 75.8% ↗ | -4 pt | Qwen3-Next 80B-A3B Instruct |
| LiveCodeBench v6 (25.02-25.05) | 55.2% ↗ | 56.6% ↗ | -1.4 pt | Qwen3-Next 80B-A3B Instruct |
| MMLU-Redux | 86.1% ↗ | 90.9% ↗ | -4.8 pt | Qwen3-Next 80B-A3B Instruct |
| SuperGPQA | 47.8% ↗ | 58.8% ↗ | -11 pt | Qwen3-Next 80B-A3B Instruct |
| TAU2-Airline | 58% ↗ | 45.5% ↗ | +12.5 pt | Qwen3 4B Thinking 2507 |
| TAU2-Retail | 53.5% ↗ | 57.3% ↗ | -3.8 pt | Qwen3-Next 80B-A3B Instruct |
| Only Qwen3 4B Thinking 2507 reports · 2 | ||||
| OJBench | 17.9% ↗ | not reported | — | — |
| TAU2-Telecom | 27.2% ↗ | not reported | — | — |
| Only Qwen3-Next 80B-A3B Instruct reports · 3 | ||||
| LMArena Elo | not reported | 1399 ↗ | — | — |
| Aider-Polyglot | not reported | 49.8% ↗ | — | — |
| MultiPL-E | not reported | 87.8% ↗ | — | — |
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 4B Thinking 2507 minus Qwen3-Next 80B-A3B 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
Official pricing pages and model cards| Spec | Qwen3 4B Thinking 2507 | Qwen3-Next 80B-A3B Instruct | Edge |
|---|---|---|---|
| Context window | 262K tokens | 262K tokens | Tie |
| Max output | — | — | — |
| Input price / 1M | — | $0.15 ↗ | — |
| Output price / 1M | — | $1.2 ↗ | — |
| Cached input / 1M | — | — | — |
| Input + output / 1M Lower is cheaper. List prices; batch, tool and regional fees excluded. | — | $1.35 | — |
| Modalities | text | text | Tie |
| Released | 2025-08-06 | 2025-09-11 | — |
| Cited benchmark scores | 15 | 16 | — |
Reliability
Qwen status
Live from /statusMore matchups
Qwen3 4B Thinking 2507 vs …
Models sharing the most benchmarksMore matchups
Qwen3-Next 80B-A3B Instruct vs …
Models sharing the most benchmarksBuilt by Respan
Which one wins on your data?
Public benchmarks are a starting point. Run Qwen3 4B Thinking 2507 and Qwen3-Next 80B-A3B Instruct on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.