Model Comparison
Granite 4.0 H Small
vs. Medium 3.1
Comparing 2 AI models · 10 benchmarks · IBM, Mistral
Recommended Pick
Mi Medium 3.1 16 metric wins Strongest on: TTFT, Latency, Reasoning
Best Value
Granite 4.0 H Small
100.0 value score
16.7 reasoning / $0.11/1M
Lowest Price
Granite 4.0 H Small
$0.06/1M input price
Best Reasoning
Medium 3.1
32.2 reasoning score
Blends available reasoning benchmarks
Best for Coding
Medium 3.1
18.3 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
Granite 4.0 H Small has the strongest quality-to-price mix at 100.0 out of 100 value points.
Price gap
Granite 4.0 H Small is 6.7x cheaper on input tokens than Medium 3.1.
Speed gap
Granite 4.0 H Small generates about 4.3x as many tokens per second as Medium 3.1.
Reasoning gap
Medium 3.1 leads Granite 4.0 H Small by 15.5 points on reasoning.
Coding gap
Medium 3.1 leads Granite 4.0 H Small by 9.8 points on coding.
Response Face-Off
Run one prompt through the selected models and compare response quality with live speed and cost context.
Granite 4.0 H Small
IBM
TTFT
—
Time
—
tok/s
—
Tokens
—
Cost
—
Medium 3.1
Mistral
TTFT
—
Time
—
tok/s
—
Tokens
—
Cost
—
Which answer was more useful?
Full Comparison
| Metric | IB Granite 4.0 H Small | Top Pick Mi Medium 3.1 |
|---|---|---|
| Pricing per 1M tokens | ||
| Input Cost | $0.06/1M | $0.40/1M |
| Output Cost | $0.25/1M | $2.00/1M |
| Blended (3:1) | $0.11/1M | $0.80/1M |
| Specifications | ||
| Organization | IBM | Mistral |
| Release Date | Sep 22, 2025 | Aug 12, 2025 |
| Performance & Speed | ||
| Throughput | 381.3 tok/s | 88.5 tok/s |
| TTFT | 9110ms | 519ms |
| Latency | 9110ms | 519ms |
| Composite Indices | ||
| Value Score | 100.0 | 25.9 |
| Reasoning Score | 16.7 | 32.2 |
| Intelligence | 10.8 | 21.3 |
| Coding | 8.5 | 18.3 |
| Math | 13.7 | 38.3 |
| Standard Benchmarks | ||
| GPQA | 41.6% | 58.8% |
| MMLU Pro | 62.4% | 68.3% |
| HLE | 3.7% | 4.4% |
| LiveCodeBench | 25.1% | 40.6% |
| AIME 2025 | 13.7% | 38.3% |
| SciCode | 20.9% | 33.8% |
| LCR | 9.0% | 19.7% |
| IFBench | 31.5% | 39.8% |
| TAU-bench v2 | 17.3% | 40.6% |
| TerminalBench Hard | 2.3% | 10.6% |
Key Takeaways
Granite 4.0 H Small offers the best value at $0.06/1M, making it ideal for high-volume applications and cost-conscious projects.
Medium 3.1 has the strongest reasoning profile with a 32.2 reasoning score, combining the available reasoning-heavy benchmarks.
Medium 3.1 reaches a 18.3 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
Granite 4.0 H Small
- Cost-sensitive applications
- High-volume processing
Medium 3.1
- Complex reasoning tasks
- Research & analysis
- Code generation
- Software development
