Gemini 3.1 Pro
Google DeepMind · released
The newest Pro-tier Gemini that developers can actually call: a heavier reasoning model for hard problem solving and multimodal analysis. It predates the current agentic benchmarks and now trails the newer Flash models on most of them.
Best for: Multimodal analysis · Audio and video input · Hard single-shot reasoning
- Intelligence Index
- 29.7
- #25 of 38
- Price per 1M tokens
- $2 in $12 out
- Blended $4.5 at 3:1
- Context window
- 1M
- Up to 66K output
- Output speed
- 122 tokens/s
- $1.3 measured cost per task
Capability profile
One 0–100 score per area, placing this model between the weakest and strongest tracked model across that area's benchmarks.
- Overall44 · #25 of 38
Based on 3 of 3 benchmarks
- Coding25 · #26 of 39
Based on 5 of 5 benchmarks
- Agents & knowledge work16 · #34 of 40
Based on 6 of 6 benchmarks
- Reasoning & science64 · #13 of 39
Based on 2 of 2 benchmarks
- Knowledge & reliability85 · #7 of 39
Based on 2 of 2 benchmarks
- Vision81 · #8 of 31
Based on 2 of 2 benchmarks
Benchmark results
All third-party results on record. Rank is among the tracked models that have been tested on that benchmark.
Overall
#25 of 38· Artificial Analysis
- Vals Index33.4%
#24 of 26· High effort· $2.18 per test· Vals AI
- Arena · Text1487 ±3
#7 of 33· 121,806 votes· Arena
Coding
#25 of 38· Artificial Analysis
- SciCode58.7%
#10 of 38· Artificial Analysis
- Arena · WebDev1447 ±5
#28 of 34· 24,296 votes· Arena
- Vibe Code Bench32.0%
#24 of 26· High effort· Vals AI
- Code Migration17.9%
#23 of 26· High effort· Vals AI
Agents & knowledge work
- GDPval-AA14.7
#32 of 39· Artificial Analysis
- Arena · Agent-7.7 ±1
#26 of 29· Arena
- Finance Agent v243.0%
#25 of 26· High effort· Vals AI
#23 of 26· High effort· Vals AI
- Legal Research Bench20.7%
#26 of 26· High effort· Vals AI
- Tax Agent Bench37.8%
#26 of 26· High effort· Vals AI
Reasoning & science
- Humanity's Last Exam47.0%
#10 of 39· Artificial Analysis
- CritPt17.7%
#17 of 39· Artificial Analysis
- ARC-AGI-2saturating77.1%
#7 of 16· Preview effort· $0.96 per task· ARC Prize
- ARC-AGI-3Not tested
- GPQA Diamondsaturating94.1%
#3 of 27· Artificial Analysis
Knowledge & reliability
#6 of 39· Artificial Analysis
- AA-LCR82.0%
#15 of 39· Artificial Analysis
Vision
- MMMU-Pro82.4%
#6 of 27· Artificial Analysis
- Arena · Vision1279 ±5
#7 of 19· 45,558 votes· Arena
Reported by Google DeepMind
Numbers the lab published at launch, run on its own harness and settings. Useful context, but not independently verified and never used in the rankings on this site.
- ARC-AGI-2 · ARC Prize verified77.1%
- GPQA Diamond · no tools94.3%
- SWE-bench Pro · single attempt54.2%
- MMMU-Pro · no tools80.5%
- Humanity's Last Exam · no tools44.4%
Specifications
- Released
- Feb 19, 2026
- Availability
- Preview
- Context window
- 1,048,576 tokens
- Max output
- 65,536 tokens
- Accepts
- Text, Images, Audio, Video, PDFs
- Knowledge cutoff
- Jan 2025
- Weights
- Closed
- Reasoning control
- Thinking level low to high
- API model ID
gemini-3.1-pro-preview
Still labelled Preview on the Gemini API, eight months after launch. It remains the newest Pro-tier Gemini that developers can call.
Pricing
- Input / 1M tokens
- $2
- Cached input / 1M
- $0.20
- Output / 1M tokens
- $12
- Blended (3:1)
- $4.5
- Measured cost / task
- $1.3
Prompts over 200K tokens cost $4 in / $18 out. Batch is half price.
Official sources
Closest alternatives
The models nearest to this one on overall score.
23 benchmarks tracked. See methodology and sources.