Gemini 4 Argon
Google·Released Sep 30, 2026
Updated Oct 8, 11:55 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
8 results · 75.4- Terminal-Bench 4.0 (AA run)Artificial Analysis · high effort57.1%Reading 85.6
- Code MigrationVals AI · high effort68.2%Reading 82.7
- Terminal-Bench 4.0Vals AI · high effort57.6%Reading 82.5
- SciCodeArtificial Analysis · high effort61.8%Reading 82.3
- Vibe Code BenchVals AI · high effort91.9%Reading 76.8
- CyberBench PatchVals AI · high effort85.7%Reading 68.5
- ProgramBenchVals AI · high effortNear the floor2.5%Reading at most 65.7
- APEX-SWEMercor · high effort49.2%Reading 65.7
Research and reasoning
4 results · 83.1- ProofBenchVals AI · high effortNear the ceiling99.0%Reading at least 89.2
- Terminal-Bench ScienceVals AI · high effort44.3%Reading 84.5
- MysteryMechanismVals AI · high effort45.5%Reading 81.3
- Humanity's Last Exam (AA run)Artificial Analysis · high effort57.1%Reading 79.2
Professional work
7 results · 88.6- Harvey Legal Agent BenchmarkVals AI · high effort19.6%Reading 103.2
- APEX-AgentsMercor · high effort82.4%Reading 102.8
- MedCodeVals AI · high effort58.8%Reading 97.5
- Finance AgentVals AI · high effort65.4%Reading 84.7
- Tax Agent BenchVals AI · high effort44.6%Reading 82.0
- Legal Research BenchVals AI · high effort54.8%Reading 78.4
- EMBVals AI · high effort75.2%Reading 76.2
Knowledge and accuracy
1 result · 59.0- AA-LCRArtificial Analysis · high effort79.7%Reading 48.5
Human preference
1 result · 68.5- Arena TextLMArena · high effort1,525.2Reading 62.8
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.
- IOIVals AI · CodingReference only100.0%
- SRE BenchVals AI · CodingWatching44.3%
- BioMysteryBenchVals AI · Research and reasoningWatching76.3%
- LegalBenchVals AI · Professional workReference only88.3%
- MedScribeVals AI · Professional workWatching87.4%
- AA Intelligence IndexArtificial Analysis · Composite indicesReference only52.6
- Vals IndexVals AI · Composite indicesReference only68.9%
- Arena WebDevLMArena · Writing and designReference only1,678.4
No result yet
Scored evaluations this model has not taken. A missing result neither adds nor subtracts.
- LiveBench Coding · LiveBench
- Vibe Code Bench 1–100 · Vals AI
- FrontierMath Tiers 1–3 · Epoch AI
- FrontierMath Tier 4 · Epoch AI
- LiveBench Reasoning · LiveBench
- Chess Puzzles · Epoch AI
- Mystery Game Puzzles · Epoch AI
- LiveBench Data Analysis · LiveBench
- τ²-Bench Telecom (AA run) · Artificial Analysis
- τ-Bench Banking (AA run) · Artificial Analysis
- SimpleQA Verified · Epoch AI
- LiveBench Language · LiveBench
- LiveBench Instruction Following · LiveBench
- BullshitBench · BullshitBench
- IFBench (AA run) · Artificial Analysis