gpt.college

Google · Gemini · Reviewed 2026-06-12

Gemini 3.1 Pro

Google DeepMind's Gemini 3.1 Pro preview, with official benchmark coverage across reasoning, coding, multimodal, tool-use, and long-context tasks.

All models
Overall · 0–100—
Status
preview
Released
—
Context
Not disclosed
Parameters
Undisclosed
Inputs
Text · Image · Audio · Video · File
Type
multimodal
Max output
— tokens
API model ID
—
Reasoning
—
Knowledge cutoff
—
Training cutoff
—
Weights / license
—
Price / 1M tokens
—
Cached input
—
High GPQA Diamond and ARC-AGI-2 resultsStrong multimodal and agentic benchmark coverage1M input token context
CodingReasoningResearchMultimodalAgentic

Independent benchmark scores

No snapshot score

Coding

—

No comparable score for this task lens in the selected independent snapshot.

No snapshot score

Reasoning

—

No comparable score for this task lens in the selected independent snapshot.

No snapshot score

Research

—

No comparable score for this task lens in the selected independent snapshot.

No snapshot score

Multimodal

—

No comparable score for this task lens in the selected independent snapshot.

No snapshot score

Agentic

—

No comparable score for this task lens in the selected independent snapshot.

Raw benchmark results

Source-backed rows with the original evaluation context preserved.

8 results
BenchmarkScoreSourceEvaluation notes
ARC-AGI-2Checked 2026-06-12 77.1 Google Verified ARC-AGI-2 score reported in Google's Gemini 3.1 Pro announcement.
BrowseCompChecked 2026-06-11 85.9% Google DeepMind BrowseComp score with search, Python, and Browse from Google DeepMind's Gemini 3.1 Pro table.
GPQA / GPQA DiamondChecked 2026-06-11 94.3% Google DeepMind GPQA Diamond no-tools score from Google DeepMind's Gemini 3.1 Pro benchmark table.
Humanity's Last Exam (HLE)Checked 2026-06-12 47.31% Benchmark project Scale Labs HLE Text Only score for gemini-3.1-pro-preview (thinking high), covering the 86% text-only HLE subset.
MMMU-ProChecked 2026-06-11 80.5% Google DeepMind MMMU-Pro no-tools score from Google DeepMind's Gemini 3.1 Pro benchmark table.
SWE-bench ProChecked 2026-06-11 54.2% Google DeepMind SWE-Bench Pro public single-attempt score from Google DeepMind's Gemini 3.1 Pro page.
SWE-bench VerifiedChecked 2026-06-11 80.6% Google DeepMind SWE-Bench Verified single-attempt score from Google DeepMind's Gemini 3.1 Pro benchmark table.
Terminal-Bench 2.0Checked 2026-06-11 68.5% Google DeepMind Terminal-Bench 2.0 score using the Terminus-2 harness from Google DeepMind's table.

Strengths

  • High GPQA Diamond and ARC-AGI-2 results
  • Strong multimodal and agentic benchmark coverage
  • 1M input token context

Limitations

  • Preview status and methodology settings should be preserved when comparing results

Model evidence · official · Checked 2026-06-10

Gemini models documentation

Use for model availability and context notes; verify before publishing claims.

Model evidence · official · Checked 2026-06-11

Gemini 3.1 Pro model page

Official Google DeepMind model page with benchmark table and methodology link for Gemini 3.1 Pro.

Model evidence · official · Checked 2026-06-12

Gemini 3.1 Pro announcement

Official Google announcement with ARC-AGI-2 verified score and release context for Gemini 3.1 Pro.

Task lenses preserve raw scores. Radar axes use score / current leader × 100; overall is the equal-weight average of all five indices. Missing evidence is not a zero or a low rank.