Qwen3.5 397B A17B
Alibaba·Released Feb 13, 2026Open weights
Updated Oct 8, 9:18 PM ET
The score is on the board's scale (reference average 50, about 15 points per standard deviation), not a percentage.
By domain
Each domain is scored on the same scale as the overall score. Coverage is the share of the domain's kinds of evaluation with a result; it says how much evidence there is, not how good the model is.
Every result behind the score
The result as the evaluator publishes it, and the reading it gives on the board's scale. A hard evaluation can give a high reading for a modest result; results near 95% or 5% only say "at least" or "at most".
Coding
3 results · 33.4- SciCodeArtificial Analysis · thinking effort44.8%Reading 35.5
- Terminal-Bench 4.0 (AA run)Artificial Analysis · thinking effortNear the floor0.0%Reading at most 34.5
- APEX-SWEMercor25.1%Reading 28.0
Research and reasoning
4 results · 29.5- Humanity's Last Exam (AA run)Artificial Analysis · thinking effort29.0%Reading 40.7
- Mystery Game PuzzlesEpoch AI · none effort18.0%Reading 34.6
- FrontierMath Tiers 1–3Epoch AI29.5%Reading 26.3
- Chess PuzzlesEpoch AI12.0%Reading 12.3
Professional work
3 results · 32.8- τ²-Bench Telecom (AA run)Artificial Analysis · thinking effortNear the ceiling95.6%Reading at least 60.0
- APEX-AgentsMercor24.9%Reading 24.2
- τ-Bench Banking (AA run)Artificial Analysis · thinking effort13.4%Reading 17.7
Knowledge and accuracy
3 results · 57.5- BullshitBenchBullshitBench · high effortCapped52.7%Reading 89.0
- IFBench (AA run)Artificial Analysis · thinking effort78.8%Reading 62.1
- AA-LCRArtificial Analysis · thinking effort77.3%Reading 44.8
Human preference
1 result · 38.1- Arena TextLMArena1,441.8Reading 38.1
Shown, not scored
Composite indices, saturated or older evaluations, evaluations still being checked, and vision, writing and multilingual results. They do not move the score.
- OTIS Mock AIMEEpoch AI · Research and reasoningReference only88.9%
- APEX-AccountingMercor · Professional workWatching0.0%
- GPQA DiamondEpoch AI · Knowledge and accuracyReference only85.9%
- AA Intelligence IndexArtificial Analysis · Composite indicesReference only18.4
- Arena VisionLMArena · VisionReference only1,246.9
- Arena WebDevLMArena · Writing and designReference only1,399.8
No result yet
Scored evaluations this model has not taken. A missing result neither adds nor subtracts.
- Terminal-Bench 4.0 · Vals AI
- LiveBench Coding · LiveBench
- Code Migration · Vals AI
- ProgramBench · Vals AI
- Vibe Code Bench · Vals AI
- Vibe Code Bench 1–100 · Vals AI
- CyberBench Patch · Vals AI
- FrontierMath Tier 4 · Epoch AI
- LiveBench Reasoning · LiveBench
- MysteryMechanism · Vals AI
- Terminal-Bench Science · Vals AI
- ProofBench · Vals AI
- Finance Agent · Vals AI
- Harvey Legal Agent Benchmark · Vals AI
- MedCode · Vals AI
- Legal Research Bench · Vals AI
- Tax Agent Bench · Vals AI
- EMB · Vals AI
- LiveBench Data Analysis · LiveBench
- SimpleQA Verified · Epoch AI
- LiveBench Language · LiveBench
- LiveBench Instruction Following · LiveBench