vs
4:9
overall capability wins (0 ties across 13 scored). Qwen3.8 27B leads overall.
- Reasoning2:4
- Coding0:3
- Arena1:0
- Tool use0:1
- Specs1:1
Verdict
Qwen3.8 27B leads coding (widest gap: +132.8 on WebDev Elo); Qwen3.6 Plus ranks higher on LMArena (+6 Elo); Qwen3.6 Plus offers ~3.8x larger context (1M vs 262K).
- Qwen3.8 27B leads coding (widest gap: +132.8 on WebDev Elo)
- Qwen3.6 Plus ranks higher on LMArena (+6 Elo)
- Qwen3.6 Plus offers ~3.8x larger context (1M vs 262K)
- Qwen3.8 27B allows ~2.0x more max output (131K vs 66K)
- Qwen3.8 27B ships open weights (Apache 2.0)
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Qwen3.6 Plus
- $1.25
- Qwen3.8 27B
- -
Price per 1M tokens
Full comparison
Organization
Alibaba
Alibaba
Family
Qwen
Qwen
License
Proprietary
Apache 2.0
Open weights
No
Yes
Release
Apr 2, 2026
Aug 14, 2026
Knowledge cutoff
-
-
API / provider
Alibaba Cloud
No primary API price listed
Modalities
text, image, video → text
text, image, video → text
Specs
Context window
1M
262K
Max output
66K
131K
Parameters
—
27B
Pricing
Input $/1M
$0.50
—
Output $/1M
$3
—
Blended $/1M (3∶1)
$1.13
-
Speed
tok/s
70
-
TTFT (s)
0.45
-
Reasoning
MMLU-Pro
—
—
GPQA Diamond
88.2
90.5
Humanity's Last Exam
27.8
33.9
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
26.4
15.6
AA-LCR v1.1
78.3
82.0
CritPt
2.9
5.4
MMMU-Pro
78.0
76.3
IFBench
75.2
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
61.7
SWE-bench Multilingual
—
—
LiveCodeBench
—
90.3
Terminal-Bench 2.1
61.4
79.8
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
40.7
46.6
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
—
42.3
DeepSWE
—
42.2
WebDev Arena
1,461
1,593
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
1,443
1,437
| Metric | Qwen3.6 Plus | Qwen3.8 27B | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Alibaba | Alibaba | — |
| Family | Qwen | Qwen | — |
| License | Proprietary | Apache 2.0 | — |
| Open weights | No | Yes | — |
| Release | Apr 2, 2026 | Aug 14, 2026 | — |
| Knowledge cutoff | - | - | — |
| API / provider | Alibaba Cloud | No primary API price listed | — |
| Modalities | text, image, video → text | text, image, video → text | — |
| Specs | |||
| Context window | 1M | 262K | A +738K |
| Max output | 66K | 131K | B +66K |
| Parameters | — | 27B | — |
| Pricing | |||
| Input $/1M | $0.50 | — | — |
| Output $/1M | $3 | — | — |
| Blended $/1M (3∶1) | $1.13 | - | — |
| Speed | |||
| tok/s | 70 | - | — |
| TTFT (s) | 0.45 | - | — |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | 88.2 | 90.5 | B +2.3 pts |
| Humanity's Last Exam | 27.8 | 33.9 | B +6.1 pts |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 26.4 | 15.6 | A +10.8 pts |
| AA-LCR v1.1 | 78.3 | 82.0 | B +3.7 pts |
| CritPt | 2.9 | 5.4 | B +2.5 pts |
| MMMU-Pro | 78.0 | 76.3 | A +1.7 pts |
| IFBench | 75.2 | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | 61.7 | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | — | 90.3 | — |
| Terminal-Bench 2.1 | 61.4 | 79.8 | B +18.4 pts |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | 40.7 | 46.6 | B +5.9 pts |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | 42.3 | — |
| DeepSWE | — | 42.2 | — |
| WebDev Arena | 1,461 | 1,593 | B +133 Elo |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | 1,443 | 1,437 | A +6 Elo |
Qwen3.6 Plus
- Context
- 1M / 66K out
- Parameters
- —
- Price
- $0.50 / $3
- Speed
- 70 tok/s · 0.45s TTFT
- Modalities
- text, image, video → text
- License
- Proprietary
Alibaba hosted flagship (Apr 2026) for agentic coding and multimodality — 1M context via DashScope; always-on reasoning.
Qwen3.8 27B
- Context
- 262K / 131K out
- Parameters
- 27B
- Price
- —
- Speed
- —
- Modalities
- text, image, video → text
- License
- Apache 2.0
Open-weight Qwen3.8 27B dense vision-language model (Apache-2.0, released Aug 14, 2026) — native image and video understanding, 262K context extensible to 1M, flexible thinking control via reasoning_effort, and strong agentic coding for its size (Terminal-Bench 2.1 73.0, SWE-bench Pro 61.7, GPQA Diamond 89.2).
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.