vs
0:10
overall capability wins (1 ties across 11 scored). Qwen3.8 Flash Next leads overall.
- Reasoning0:6
- Coding0:2
- Tool use0:1
- Composite indices0:2
- Specs0:1
Verdict
Qwen3.8 Flash Next leads coding (widest gap: +41.2 on TermBench); Qwen3.8 Flash Next allows ~1.6x more max output (131K vs 82K).
- Qwen3.8 Flash Next leads coding (widest gap: +41.2 on TermBench)
- Qwen3.8 Flash Next allows ~1.6x more max output (131K vs 82K)
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Qwen3.6 35B A3B
- -
- Qwen3.8 Flash Next
- $0.27
Price per 1M tokens
Full comparison
Organization
Alibaba
Alibaba
Family
Qwen
Qwen
License
Apache 2.0
Qwen Community License 1.0
Open weights
Yes
Yes
Release
Apr 16, 2026
Aug 26, 2026
Knowledge cutoff
-
-
API / provider
No primary API price listed
Alibaba (QwenCloud)
Modalities
text, image, video → text
text, image, video → text
Specs
Context window
262K
262K
Max output
82K
131K
Parameters
35B (3B act.)
—
Pricing
Input $/1M
—
$0.15
Output $/1M
—
$0.47
Blended $/1M (3∶1)
-
$0.23
Speed
tok/s
-
77.3
TTFT (s)
-
2.79
Reasoning
MMLU-Pro
—
—
GPQA Diamond
84.1
92.3
Humanity's Last Exam
22.2
38.0
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
18.8
24.5
AA-LCR v1.1
71.7
79.7
CritPt
0.3
11.1
MMMU-Pro
75.0
79.8
IFBench
64.4
—
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
—
—
SWE-bench Multilingual
—
—
LiveCodeBench
—
—
Terminal-Bench 2.1
44.9
86.1
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
36.6
50.6
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
—
—
DeepSWE
—
—
WebDev Arena
—
1,635
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
—
—
| Metric | Qwen3.6 35B A3B | Qwen3.8 Flash Next | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Alibaba | Alibaba | — |
| Family | Qwen | Qwen | — |
| License | Apache 2.0 | Qwen Community License 1.0 | — |
| Open weights | Yes | Yes | — |
| Release | Apr 16, 2026 | Aug 26, 2026 | — |
| Knowledge cutoff | - | - | — |
| API / provider | No primary API price listed | Alibaba (QwenCloud) | — |
| Modalities | text, image, video → text | text, image, video → text | — |
| Specs | |||
| Context window | 262K | 262K | tie |
| Max output | 82K | 131K | B +49K |
| Parameters | 35B (3B act.) | — | — |
| Pricing | |||
| Input $/1M | — | $0.15 | — |
| Output $/1M | — | $0.47 | — |
| Blended $/1M (3∶1) | - | $0.23 | — |
| Speed | |||
| tok/s | - | 77.3 | — |
| TTFT (s) | - | 2.79 | — |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | 84.1 | 92.3 | B +8.2 pts |
| Humanity's Last Exam | 22.2 | 38.0 | B +15.8 pts |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 18.8 | 24.5 | B +5.7 pts |
| AA-LCR v1.1 | 71.7 | 79.7 | B +8.0 pts |
| CritPt | 0.3 | 11.1 | B +10.8 pts |
| MMMU-Pro | 75.0 | 79.8 | B +4.8 pts |
| IFBench | 64.4 | — | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | — | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | — | — | — |
| Terminal-Bench 2.1 | 44.9 | 86.1 | B +41.2 pts |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | 36.6 | 50.6 | B +14.0 pts |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | — | — | — |
| DeepSWE | — | — | — |
| WebDev Arena | — | 1,635 | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | — | — | — |
Qwen3.6 35B A3B
- Context
- 262K / 82K out
- Parameters
- 35B (3B act.)
- Price
- —
- Speed
- —
- Modalities
- text, image, video → text
- License
- Apache 2.0
Efficient Qwen3.6 MoE sibling to Qwen3.6-27B (35B total / 3B active) — same 262K native context and multimodal input, Apache 2.0.
Qwen3.8 Flash Next
- Context
- 262K / 131K out
- Parameters
- —
- Price
- $0.15 / $0.47
- Speed
- 77.3 tok/s · 2.79s TTFT
- Modalities
- text, image, video → text
- License
- Qwen Community License 1.0
Open-weight sparse-MoE Flash tier (6B active) previewing the Qwen4 architecture, released Aug 26, 2026 — 262K native context extensible to ~1M with YaRN, tuned for cost efficiency at $0.15/$0.47 per 1M (cache-hit $0.016). Reported total parameter count varies by source (~125-180B). Benchmark figures are Alibaba's self-reported launch numbers (SWE-bench Pro 62.5, HLE 35.9); Artificial Analysis places it at 56 on its Intelligence Index.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.