Gemini 4 Argon

Leaderboards

Where Gemini 4 Argon sits among the top 20 in each category.

By task

Scored from votes on each kind of prompt alone, on the same 1,000-point scale.

Test results

Every test Gemini 4 Argon has a published result on, out of 100%, with its rank among the models that took it.

• DeepSWE 1.1

  #5 / 50

  DeepSWE 1.1 evaluates software engineering agents using the mini-swe-agent harness.

  77.9%

  Self-reported · In the coding score

• LVBench

  #1 / 31

  LVBench is an extreme long video understanding benchmark designed to evaluate multimodal models on videos up to two hours in duration. It contains 6 major categories and 21 subcategories, with videos averaging five times longer than existing datasets. The benchmark addresses applications requiring comprehension of extremely long videos.

  91.7%

  Self-reported

• Terminal-Bench 4.0

  #5 / 28

  Terminal-Bench 4.0 evaluates AI agents on real-world work in terminal and command-line containerized environments across 66 tasks, with emphasis on science-adjacent and frontier engineering problems. Relative to earlier Terminal-Bench releases, 4.0 increases timeouts and adaptively raises RAM/CPU on selected tasks to reduce harness and resource confounds.

  57.4%

  Self-reported · In the coding score· In the agents score

• Agents' Last Exam

  #12 / 24

  Agents' Last Exam is a challenging benchmark for AI agents on hard, long-horizon tasks that test sustained reasoning, planning, and tool use, reported with and without tool access.

  39.5%

  Self-reported · In the agents score

• AutomationBench

  #4 / 24

  AutomationBench is a tool-use benchmark that evaluates AI agents on automating real-world workflows, testing their ability to orchestrate tools and complete multi-step automation tasks.

  51.3%

  Self-reported · In the agents score

• OSWorld 2.0

  #6 / 17

  OSWorld 2.0 is a benchmark of 108 long-horizon, real-world computer-use workflows spanning everyday and professional tasks. Each task is an end-to-end workflow that takes human users a median of about 1.6 hours, scored with a binary-completion metric, and targets challenges such as dynamic environments, cross-source reasoning, and implicit-state inference.

  69.2%

  Self-reported

• Graphwalks BFS <128k

  #1 / 13

  A graph reasoning benchmark that evaluates language models' ability to perform breadth-first search (BFS) operations on graphs with context length under 128k tokens, returning nodes reachable at specified depths.

  99.7%

  Self-reported

• Graphwalks BFS >128k

  #2 / 12

  A graph reasoning benchmark that evaluates language models' ability to perform breadth-first search (BFS) operations on graphs with context length over 128k tokens, testing long-context reasoning capabilities.

  84.2%

  Self-reported

• PostTrainBench

  #1 / 9

  PostTrainBench evaluates a model's ability to autonomously post-train base models. Given pretrain-only base models, the agent must complete the full pipeline of data synthesis, training, evaluation, and iteration within a time budget, scored across downstream benchmarks such as AIME2025, BFCL, GPQA Main, GSM8K, and HumanEval.

  45.3%

  Self-reported

• Terminal-Bench-Science 0.1

  #4 / 7

  Terminal-Bench-Science 0.1 is a Stanford-led community benchmark of 70 tasks drawn from scientific research workflows across the life, physical, earth, mathematical, and engineering sciences. Tasks are authored and reviewed by scientists and researchers; scores are typically reported as accuracy with relatively large standard error due to strongly bimodal per-task outcomes.

  57.6%

  Self-reported

• Vals Finance Agent v2

  #1 / 6

  Vals Finance Agent v2 evaluates AI agents on realistic financial analyst tasks.

  65.4%

  Self-reported

• FrontierSWE V2

  #3 / 5

  FrontierSWE V2 evaluates agents on open-ended technical projects and reports mean@5. Kept separate from FrontierSWE V1 dominance-score results, which are not comparable.

  55.0%

  Self-reported

• Harvey's Legal Agent Benchmark

  #1 / 4

  Harvey's Legal Agent Benchmark evaluates AI agents on complex legal workflows and reports the all-pass rate.

  19.6%

  Self-reported

• Chartography (no tools)

  #1 / 3

  Chartography visual chart recognition evaluated without tools, distinct from tool-assisted results.

  71.6%

  Self-reported

• LABBench2

  #1 / 3

  LABBench2 evaluates models on real-world biology research tasks.

  88.8%

  Self-reported

• CWE-Bench

  #1 / 2

  CWE-Bench evaluates autonomous agents on finding and patching real-world software vulnerabilities.

  68.0%

  Unverified

• RiemannBench

  #1 / 1

  RiemannBench evaluates models on advanced mathematics and scientific reasoning tasks, as reported on the Surge public leaderboard.

  76.0%

  Unverified

• Vals Index

  #1 / 1

  Vals Index measures economic impact across finance, coding, legal, and tax work, with every sector weighted by its contribution to U.S. GDP.

  68.9%

  Unverified

• Vibe Code Bench

  #1 / 1

  Vibe Code Bench is a Vals AI agentic coding benchmark evaluating models on end-to-end coding workflows.

  91.9%

  Unverified