Model Comparison
Llama 3.2 Instruct 90B (Vision)
vs. Gemini 3.1 Pro Preview
Comparing 2 AI models · 11 benchmarks · Meta, Google
Recommended Pick
Go Gemini 3.1 Pro Preview 6 metric wins Strongest on: Throughput, Reasoning, Intelligence
Best Value
Llama 3.2 Instruct 90B (Vision)
100.0 value score
25.6 reasoning / $1.38/1M
Lowest Price
Llama 3.2 Instruct 90B (Vision)
$1.38/1M input price
Best Reasoning
Gemini 3.1 Pro Preview
65.3 reasoning score
Blends available reasoning benchmarks
Best for Coding
Gemini 3.1 Pro Preview
55.5 coding index
Composite Indices
Higher is better; speed and price are normalized
Standard Benchmarks
Only benchmarks with data are shown
Differences That Matter
Best value
Llama 3.2 Instruct 90B (Vision) has the strongest quality-to-price mix at 100.0 out of 100 value points.
Price gap
Llama 3.2 Instruct 90B (Vision) is 1.4x cheaper on input tokens than Gemini 3.1 Pro Preview.
Speed gap
Gemini 3.1 Pro Preview generates about 3.0x as many tokens per second as Llama 3.2 Instruct 90B (Vision).
Reasoning gap
Gemini 3.1 Pro Preview leads Llama 3.2 Instruct 90B (Vision) by 39.7 points on reasoning.
Top-pick rationale
Gemini 3.1 Pro Preview wins 6 measurable categories, including Throughput, Reasoning, Intelligence, GPQA.
Response Face-Off
Run one prompt through the selected models and compare response quality with live speed and cost context.
Llama 3.2 Instruct 90B (Vision)
Meta
TTFT
—
Time
—
tok/s
—
Tokens
—
Cost
—
Gemini 3.1 Pro Preview
TTFT
—
Time
—
tok/s
—
Tokens
—
Cost
—
Which answer was more useful?
Full Comparison
| Metric | Me Llama 3.2 Instruct 90B (Vision) | Top Pick Go Gemini 3.1 Pro Preview |
|---|---|---|
| Pricing per 1M tokens | ||
| Input Cost | $1.38/1M | $2.00/1M |
| Output Cost | $1.38/1M | $12.00/1M |
| Blended (3:1) | $1.38/1M | $4.50/1M |
| Specifications | ||
| Organization | Meta | |
| Release Date | Sep 25, 2024 | Feb 19, 2026 |
| Performance & Speed | ||
| Throughput | 48.7 tok/s | 147.5 tok/s |
| TTFT | 553ms | 40202ms |
| Latency | 553ms | 40202ms |
| Composite Indices | ||
| Value Score | 100.0 | 78.3 |
| Reasoning Score | 25.6 | 65.3 |
| Intelligence | 11.9 | 57.2 |
| Coding | — | 55.5 |
| Standard Benchmarks | ||
| GPQA | 43.2% | 94.1% |
| MMLU Pro | 67.1% | — |
| HLE | 4.9% | 44.7% |
| LiveCodeBench | 21.4% | — |
| MATH 500 | 62.9% | — |
| AIME (Original) | 5.0% | — |
| SciCode | 24.0% | 58.9% |
| LCR | — | 72.7% |
| IFBench | — | 77.1% |
| TAU-bench v2 | — | 95.6% |
| TerminalBench Hard | — | 53.8% |
Key Takeaways
Llama 3.2 Instruct 90B (Vision) offers the best value at $1.38/1M, making it ideal for high-volume applications and cost-conscious projects.
Gemini 3.1 Pro Preview has the strongest reasoning profile with a 65.3 reasoning score, combining the available reasoning-heavy benchmarks.
Gemini 3.1 Pro Preview reaches a 55.5 coding index, making it the top choice for software development and code generation tasks.
All models support context windows of ∞+ tokens, suitable for processing lengthy documents and maintaining extended conversations.
When to Choose Each Model
Llama 3.2 Instruct 90B (Vision)
- Cost-sensitive applications
- High-volume processing
Gemini 3.1 Pro Preview
- Complex reasoning tasks
- Research & analysis
- Code generation
- Software development
