Compare / Qwen3-Next 80B-A3B Thinking vs Qwen3-Coder-Next
Qwen3-Next 80B-A3B Thinking vs Qwen3-Coder-Next
Updated November 2026
Verdict
Qwen3-Next 80B-A3B Thinking and Qwen3-Coder-Next are roughly tied on the Small LM Index (63.7 vs 63.1). On the 2 benchmarks both labs report, Qwen3-Next 80B-A3B Thinking wins 2 and Qwen3-Coder-Next wins 0. Both need about 50.8 GB of memory at 4-bit quantization. Only Qwen3-Coder-Next has a hosted API on OpenRouter ($0.29 per million tokens); the other has to be self-hosted.
Pick Qwen3-Next 80B-A3B Thinking if…
- you need precise instruction following
Pick Qwen3-Coder-Next if…
- you want a hosted API rather than running it yourself
- you want the more widely used model (more quantized builds and community support)
Specifications
| Qwen3-Next 80B-A3B Thinking | Qwen3-Coder-Next | |
|---|---|---|
| Small LM Index | 63.7 (likely 62–66) | 63.1 (likely 54–70) |
| Lab | Alibaba (Qwen) · China | Alibaba (Qwen) · China |
| Released | 2025-09-11 | 2026-01-30 |
| Parameters | 80B (3B active, MoE) | 79.7B (3B active, MoE) |
| Memory at 4-bit | ~50.8 GB | ~50.6 GB |
| Context window | 262K tokens | 262K tokens |
| Input | text | text |
| Reasoning | Always thinks | No thinking |
| License | Apache-2.0 | Apache-2.0 |
| API price / 1M tokens | Self-host only | $0.29 ($0.12 in, $0.80 out) |
| Hugging Face downloads (30 days) | 43.9K | 430.1K |
Benchmarks
Higher is better. Bold marks the winner where both labs report a score. ≈ marks the index model's estimate for a score that wasn't reported, ✓ an independent measurement and * a score measured by another lab.
| Benchmark | Qwen3-Next 80B-A3B Thinking | Qwen3-Coder-Next |
|---|---|---|
| MMLU-Pro Knowledge | 82.7 | ≈ 81 |
| GPQA Diamond Knowledge | 77.2 | 73.7✓ |
| Humanity's Last Exam Reasoning | ≈ 11 | 10.1✓ |
| AIME Math | 87.8 | ≈ 87 |
| LiveCodeBench Coding | 68.7 | ≈ 73 |
| SWE-bench Verified Coding | ≈ 55 | 70.6 |
| Terminal-Bench Agentic | — | 36.2 |
| τ²-Bench Agentic | — | 79.5✓ |
| IFBench Instruction following | 61.5 | 35.2✓ |
| AA-LCR Long context | — | 47.0✓ |
| SciCode Coding | — | 36.2✓ |
Scores are mostly reported by each lab with its own settings, so small gaps are noise. See the methodology. Open in the interactive comparison to add more models.
Questions
Which is better, Qwen3-Next 80B-A3B Thinking or Qwen3-Coder-Next?
Qwen3-Next 80B-A3B Thinking and Qwen3-Coder-Next are roughly tied on the Small LM Index (63.7 vs 63.1). On the 2 benchmarks both labs report, Qwen3-Next 80B-A3B Thinking wins 2 and Qwen3-Coder-Next wins 0.
Which is cheaper to run?
Only Qwen3-Coder-Next has a hosted API on OpenRouter ($0.29 per million tokens); the other has to be self-hosted. Both need about 50.8 GB of memory at 4-bit quantization.
Which has the longer context window?
Both have a 262K-token context window.
Can I use Qwen3-Next 80B-A3B Thinking and Qwen3-Coder-Next commercially?
Both are released under the Apache-2.0 license. Apache 2.0 and MIT allow commercial use; other licenses can add conditions, so read the license before shipping a product.