Compare
Four models, one page
Pick the models you are actually choosing between. Every row is scaled within itself, so a bar means something next to its neighbours and nothing across rows.
| Specification | Gemini 3.1 Pro PreviewGoogle | Claude Opus 5Anthropic |
|---|---|---|
| Released | Feb 19, 2026 | Jul 24, 2026 |
| Context window | 1.05M | 1M |
| Max output | 66K | 128K |
| Input / 1M | $2 | $5 |
| Output / 1M | $12 | $25 |
| Cost per task | $1.06 | $2.25 |
| Arena Elo | 1,333 | 1,393 |
| Serving providers | 2 | 5 |
| Parameters | Undisclosed | Undisclosed |
| Licence | Proprietary | Proprietary |
| Capabilities |
|
|
Metrics side by side
Each row is scaled to the largest value in that row — bars compare within a row, never across rows
- Gemini 3.1 Pro Preview
- Claude Opus 5
Intelligence Index
Gemini 3.1 Pro Preview47.7Claude Opus 563.1Coding Index
Gemini 3.1 Pro Preview68.8Claude Opus 578.0Agentic Index
Gemini 3.1 Pro Preview23.0Claude Opus 559.2Output speed
Gemini 3.1 Pro Preview110 t/sClaude Opus 587 t/sContext window
Gemini 3.1 Pro Preview1.05MClaude Opus 51MLatency · lower is better
Gemini 3.1 Pro Preview2.86sClaude Opus 51.12sBlended price / 1M · lower is better
Gemini 3.1 Pro Preview$4.5Claude Opus 5$10Cost per task · lower is better
Gemini 3.1 Pro Preview$1.06Claude Opus 5$2.25
Rows marked “lower is better” still draw a longer bar for a larger number — read the value, not just the length. Arena Elo is in the specification table above instead: it has no meaningful zero, so a bar would flatten the gaps.
View as table
| Metric | Gemini 3.1 Pro Preview | Claude Opus 5 |
|---|---|---|
| Intelligence Index | 47.7 | 63.1 |
| Coding Index | 68.8 | 78.0 |
| Agentic Index | 23.0 | 59.2 |
| Output speed | 110 t/s | 87 t/s |
| Context window | 1.05M | 1M |
| Latency · lower is better | 2.86s | 1.12s |
| Blended price / 1M · lower is better | $4.5 | $10 |
| Cost per task · lower is better | $1.06 | $2.25 |
Evaluation scores
Percentage correct on a common 0–100% scale
- Gemini 3.1 Pro Preview
- Claude Opus 5
GPQA Diamond
Gemini 3.1 Pro Preview94.1%Claude Opus 593.2%Humanity's Last Exam
Gemini 3.1 Pro Preview47.0%Claude Opus 554.9%SciCode
Gemini 3.1 Pro Preview58.9%Claude Opus 555.7%τ²-bench
Gemini 3.1 Pro Preview95.6%Claude Opus 542.1%Terminal-Bench Hard
Gemini 3.1 Pro Preview53.8%Claude Opus 5not measuredAA-LCR long context
Gemini 3.1 Pro Preview79.0%Claude Opus 575.7%
A missing bar means that evaluation was not run for that model — it is not a zero.
View as table
| Evaluation | Gemini 3.1 Pro Preview | Claude Opus 5 |
|---|---|---|
| GPQA Diamond | 94.1% | 93.2% |
| Humanity's Last Exam | 47.0% | 54.9% |
| SciCode | 58.9% | 55.7% |
| τ²-bench | 95.6% | 42.1% |
| Terminal-Bench Hard | 53.8% | — |
| LiveCodeBench | — | — |
| AA-LCR long context | 79.0% | 75.7% |
| AIME 2025 | — | — |
Estimated from list pricing: 50K input tokens plus 80K output tokens for reasoning models (25K for non-reasoning).