Private AI
Qwen3 Vision‑Language model (235B MoE, ≈22B active) tuned for instruction following and grounded visual QA. Excels at image understanding, dense OCR, charts/diagrams, and multi‑image context.
Context Window
128.0K
Max Output
262.1K
Input Price (Auto)
$0.30/1M
Output Price (Auto)
$1.20/1M
Capabilities
Performance metrics and benchmarks
Sourced from Artificial Analysis.
Intelligence Index
20.8
Auto routing is available for this model. Explicit provider selection is not available.
Loading provider options…
Coding Index
16.5
Agentic Index
19.2
GPQA Diamond
Graduate-level scientific reasoning
71.2%
Better than 57% of models compared
HLE
Humanity's Last Exam
6.3%
Better than 49% of models compared
IFBench
Instruction-following benchmark
42.7%
Better than 46% of models compared
T²-Bench Telecom
Conversational AI agents in dual-control scenarios
35.1%
Better than 44% of models compared
AA-LCR
Long context reasoning evaluation
31.7%
Better than 44% of models compared
GDPval-AA
Economically valuable tasks
11.9%
Better than 39% of models compared
CritPt
Research-level physics reasoning
0.0%
Better than 27% of models compared
SciCode
Python programming for scientific computing
35.9%
Better than 59% of models compared
Terminal-Bench Hard
Agentic coding and terminal use
6.8%
AIME 2025
American Invitational Mathematics Examination 2025
70.7%
Better than 65% of models compared
MMLU-Pro
Professional and academic subject knowledge
82.3%
Better than 80% of models compared
AA-Omniscience Accuracy
Proportion of correctly answered questions
20.2%
Better than 48% of models compared
Last updated Jun 4, 2026
Artificial AnalysisBetter than 36% of models compared
LiveCodeBench
Contamination-free coding benchmark
59.4%
Better than 66% of models compared
AA-Omniscience Hallucination Rate
Rate of incorrect answers among non-correct responses
87.0%
Better than 28% of models compared
