Shared benchmarks

18
Shared benchmarks for Gemini 3 Flash and Gemini 3.5 Flash
BenchmarkGemini 3 FlashGemini 3.5 Flash
  • 42.6%
  • 57.9%
Evaluation details for Finance Agent Benchmark 2 — October 2026 evaluation

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

  • 89.4%high
  • 83.2%
  • 88.9%low
  • 92.8%high
Evaluation details for GPQA Diamond — Epoch AI harness

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

Gemini 3 Flash

Record for Gemini 3 Flash on Graduate-Level Google-Proof Q&A Diamond — Epoch AI harness

Gemini 3.5 Flash

  • 87.9%
  • 92.7%
Evaluation details for GPQA Diamond — Vals AI harness

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

Gemini 3 Flash

Record for Gemini 3 Flash on Graduate-Level Google-Proof Q&A Diamond — Vals AI harness

Gemini 3.5 Flash

Record for Gemini 3.5 Flash on Graduate-Level Google-Proof Q&A Diamond — Vals AI harness
GPQAPotentially non-equivalent: Benchmark versions differ. Evaluator sets differ.
  • 90.4%Diamond · Google DeepMind
  • 85.7%Diamond — Mercor single-shot · Mercor
Potentially non-equivalent · Evaluation details for GPQA

Benchmark versions differ. Evaluator sets differ. Evaluation methodology is not recorded by the registry.

Gemini 3 Flash

Record for Gemini 3 Flash on Graduate-Level Google-Proof Q&A Diamond

Gemini 3.5 Flash

  • 0.0%
  • 2.5%
Evaluation details for HLAB October 2026 evaluation

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

Gemini 3 Flash

Record for Gemini 3 Flash on Harvey's Legal Agent Benchmark October 2026 evaluation

Gemini 3.5 Flash

  • 33.7%
  • 40.2%
Evaluation details for Humanity's Last Exam Full set — No tools

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

Gemini 3 Flash

Record for Gemini 3 Flash on Humanity's Last Exam Full set — No tools

Gemini 3.5 Flash

Record for Gemini 3.5 Flash on Humanity's Last Exam Full set — No tools
  • 62.0%
  • 83.6%
Evaluation details for MCP Atlas Public, April 2026 update

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

Gemini 3 Flash

Record for Gemini 3 Flash on MCP Atlas Public, April 2026 update

Gemini 3.5 Flash

  • 88.6%
  • 89.5%
Evaluation details for MMLU-Pro Original — Vals AI harness

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

Gemini 3 Flash

Gemini 3.5 Flash

  • 81.2%
  • 83.6%
Evaluation details for MMMU-Pro No tools

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

Gemini 3 Flash

Record for Gemini 3 Flash on MMMU-Pro No tools

Gemini 3.5 Flash

Record for Gemini 3.5 Flash on MMMU-Pro No tools
  • 87.6%
  • 88.3%
Evaluation details for MMMU-Pro Vals AI harness

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

Gemini 3 Flash

Record for Gemini 3 Flash on MMMU-Pro Vals AI harness

Gemini 3.5 Flash

Record for Gemini 3.5 Flash on MMMU-Pro Vals AI harness
  • 66.8%
  • 66.2%
Evaluation details for SimpleQA Verified Original — Epoch AI harness

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

  • 75.4%
  • 79.3%
Evaluation details for SWE-bench Verified — Epoch AI harness

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

Gemini 3 Flash

Record for Gemini 3 Flash on SWE-bench Verified — Epoch AI harness

Gemini 3.5 Flash

  • 75.0%
  • 78.8%
Evaluation details for SWE-bench Verified — Vals AI mini-swe-agent

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

Gemini 3 Flash

Gemini 3.5 Flash

SWE-bench ProPotentially non-equivalent: Benchmark versions differ. Evaluator sets differ.
  • 34.6%Public — Scale AI SWE-Agent · Scale AI
  • 55.1%Public · Google DeepMind
Potentially non-equivalent · Evaluation details for SWE-bench Pro

Benchmark versions differ. Evaluator sets differ. Evaluation methodology is not recorded by the registry.

Gemini 3 Flash

Record for Gemini 3 Flash on SWE-bench Pro Public — Scale AI SWE-Agent

Gemini 3.5 Flash

Record for Gemini 3.5 Flash on SWE-bench Pro Public
  • 51.7%
  • 67.4%
Evaluation details for Terminal-Bench 2.0 — Vals AI Terminus 2

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

Gemini 3 Flash

Record for Gemini 3 Flash on Terminal-Bench 2.0 — Vals AI Terminus 2

Gemini 3.5 Flash

  • 53.9%
  • 74.2%
Evaluation details for Terminal-Bench 2.1 — Vals AI Terminus 2

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

Gemini 3 Flash

Record for Gemini 3 Flash on Terminal-Bench 2.1 — Vals AI Terminus 2

Gemini 3.5 Flash

Record for Gemini 3.5 Flash on Terminal-Bench 2.1 — Vals AI Terminus 2
Toolathlon OriginalPotentially non-equivalent: Evaluator sets differ.
  • 49.4%Original · Toolathlon Authors
  • 56.5%Original · Google DeepMind
Potentially non-equivalent · Evaluation details for Toolathlon Original

Evaluator sets differ. Evaluation methodology is not recorded by the registry.

Gemini 3 Flash

Record for Gemini 3 Flash on Toolathlon Original

Gemini 3.5 Flash

Record for Gemini 3.5 Flash on Toolathlon Original
  • 20.2%
  • 48.7%
Evaluation details for Vibe Code Bench 1.1 — OpenHands, October 2026 evaluation

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.