vs
7:0
overall capability wins (0 ties across 7 scored). Qwen3.8 Max leads overall.
- Reasoning4:0
- Specs3:0
Verdict
Qwen3.8 Max edges reasoning & knowledge; Qwen3.8 Max offers ~15.3x larger context (1M vs 66K); Qwen3.8 Max allows ~8.0x more max output (131K vs 16K).
- Qwen3.8 Max edges reasoning & knowledge
- Qwen3.8 Max offers ~15.3x larger context (1M vs 66K)
- Qwen3.8 Max allows ~8.0x more max output (131K vs 16K)
Cost for 1M in + 250K out
Illustrative chat workload at primary-provider list prices.
- Qwen3.8 Max
- $3.50
- Qwen3-Omni-30B-A3B-Instruct
- -
Price per 1M tokens
Full comparison
Organization
Alibaba
Alibaba
Family
Qwen
Qwen
License
Qwen3.8-Max License
Apache 2.0
Open weights
Yes
Yes
Release
Aug 3, 2026
Sep 22, 2025
Knowledge cutoff
-
-
API / provider
Alibaba Cloud
No primary API price listed
Modalities
text, image, video → text
text, image, audio, video → text, audio
Specs
Context window
1M
66K
Max output
131K
16K
Parameters
2.4T (95B act.)
30B (3B act.)
Pricing
Input $/1M
$2
—
Output $/1M
$6
—
Blended $/1M (3∶1)
$3
-
Speed
tok/s
86
-
TTFT (s)
0.4
-
Reasoning
MMLU-Pro
—
—
GPQA Diamond
92.8
62.0
Humanity's Last Exam
43.1
4.6
AIME 2025
—
—
MATH-500
—
—
Humanity's Last Exam (with tools)
—
—
AA-Omniscience Accuracy
31.7
14.3
AA-LCR v1.1
80.3
—
CritPt
17.7
—
MMMU-Pro
82.8
55.5
IFBench
—
31.2
Chartography
—
—
Chartography (With Tools)
—
—
Coding
SWE-bench Verified
—
—
SWE-bench Pro
67.7
—
SWE-bench Multilingual
—
—
LiveCodeBench
—
—
Terminal-Bench 2.1
81.3
—
Aider Polyglot
—
—
Terminal-Bench 3
—
—
BigCodeBench
—
—
SciCode
52.1
—
CursorBench
—
—
SWE-Rebench
—
—
NL2Repo-Bench
55.9
—
DeepSWE
56.6
—
WebDev Arena
1,671
—
Terminal-Bench 4.0
—
—
CursorBench 4.0
—
—
FrontierCode v1.1 (Main)
—
—
Arena
LMArena Elo
1,481
—
| Metric | Qwen3.8 Max | Qwen3-Omni-30B-A3B-Instruct | Delta |
|---|---|---|---|
| Identity | |||
| Organization | Alibaba | Alibaba | — |
| Family | Qwen | Qwen | — |
| License | Qwen3.8-Max License | Apache 2.0 | — |
| Open weights | Yes | Yes | — |
| Release | Aug 3, 2026 | Sep 22, 2025 | — |
| Knowledge cutoff | - | - | — |
| API / provider | Alibaba Cloud | No primary API price listed | — |
| Modalities | text, image, video → text | text, image, audio, video → text, audio | — |
| Specs | |||
| Context window | 1M | 66K | A +934K |
| Max output | 131K | 16K | A +115K |
| Parameters | 2.4T (95B act.) | 30B (3B act.) | A +2K |
| Pricing | |||
| Input $/1M | $2 | — | — |
| Output $/1M | $6 | — | — |
| Blended $/1M (3∶1) | $3 | - | — |
| Speed | |||
| tok/s | 86 | - | — |
| TTFT (s) | 0.4 | - | — |
| Reasoning | |||
| MMLU-Pro | — | — | — |
| GPQA Diamond | 92.8 | 62.0 | A +30.8 pts |
| Humanity's Last Exam | 43.1 | 4.6 | A +38.5 pts |
| AIME 2025 | — | — | — |
| MATH-500 | — | — | — |
| Humanity's Last Exam (with tools) | — | — | — |
| AA-Omniscience Accuracy | 31.7 | 14.3 | A +17.4 pts |
| AA-LCR v1.1 | 80.3 | — | — |
| CritPt | 17.7 | — | — |
| MMMU-Pro | 82.8 | 55.5 | A +27.3 pts |
| IFBench | — | 31.2 | — |
| Chartography | — | — | — |
| Chartography (With Tools) | — | — | — |
| Coding | |||
| SWE-bench Verified | — | — | — |
| SWE-bench Pro | 67.7 | — | — |
| SWE-bench Multilingual | — | — | — |
| LiveCodeBench | — | — | — |
| Terminal-Bench 2.1 | 81.3 | — | — |
| Aider Polyglot | — | — | — |
| Terminal-Bench 3 | — | — | — |
| BigCodeBench | — | — | — |
| SciCode | 52.1 | — | — |
| CursorBench | — | — | — |
| SWE-Rebench | — | — | — |
| NL2Repo-Bench | 55.9 | — | — |
| DeepSWE | 56.6 | — | — |
| WebDev Arena | 1,671 | — | — |
| Terminal-Bench 4.0 | — | — | — |
| CursorBench 4.0 | — | — | — |
| FrontierCode v1.1 (Main) | — | — | — |
| Arena | |||
| LMArena Elo | 1,481 | — | — |
Qwen3.8 Max
- Context
- 1M / 131K out
- Parameters
- 2.4T (95B act.)
- Price
- $2 / $6
- Speed
- 86 tok/s · 0.4s TTFT
- Modalities
- text, image, video → text
- License
- Qwen3.8-Max License
Alibaba's Aug 2026 Max flagship (2.4T MoE / 95B active) — frontier agentic coding via QwenCloud / DashScope with 1M multimodal context and $2/$6 API pricing. Open weights released Aug 12, 2026 as Qwen3.8-2.4T-A95B (text-only, thinking-mandatory, Qwen3.8-Max license); the hosted version adds vision input, non-thinking mode, and built-in tools.
Qwen3-Omni-30B-A3B-Instruct
- Context
- 66K / 16K out
- Parameters
- 30B (3B act.)
- Price
- —
- Speed
- —
- Modalities
- text, image, audio, video → text, audio
- License
- Apache 2.0
Alibaba's first natively end-to-end omni-modal model (text/image/audio/video in; text and streamed speech out) via a Thinker-Talker MoE architecture; DashScope hosting pricing not publicly listed.
Benchmark charts
Winner bars are emphasized. Per-benchmark deltas sit above each chart.