Model Comparison
V3 (Dec '24)
vs. V3.1 Terminus (Non-reasoning)
Comparing 2 AI models · 12 benchmarks · DeepSeek
Recommended Pick
De V3.1 Terminus (Non-reasoning) 16 metric wins Strongest on: Value, Input price, Blended price
Best Value
V3.1 Terminus (Non-reasoning)
100.0 value score
43.9 reasoning / $0.45/1M
Lowest Price
V3.1 Terminus (Non-reasoning)
$0.27/1M input price
Best Reasoning
V3.1 Terminus (Non-reasoning)
43.9 reasoning score
Blends available reasoning benchmarks
Best for Coding
V3.1 Terminus (Non-reasoning)
31.9 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
V3.1 Terminus (Non-reasoning) has the strongest quality-to-price mix at 100.0 out of 100 value points.
Price gap
V3.1 Terminus (Non-reasoning) is 1.5x cheaper on input tokens than V3 (Dec '24).
Reasoning gap
V3.1 Terminus (Non-reasoning) leads V3 (Dec '24) by 9.3 points on reasoning.
Coding gap
V3.1 Terminus (Non-reasoning) leads V3 (Dec '24) by 15.5 points on coding.
Top-pick rationale
V3.1 Terminus (Non-reasoning) wins 16 measurable categories, including Value, Input price, Blended price, Reasoning.
Response Face-Off
Run one prompt through the selected models and compare response quality with live speed and cost context.
V3 (Dec '24)
DeepSeek
TTFT
—
Time
—
tok/s
—
Tokens
—
Cost
—
V3.1 Terminus (Non-reasoning)
DeepSeek
TTFT
—
Time
—
tok/s
—
Tokens
—
Cost
—
Which answer was more useful?
Full Comparison
| Metric | De V3 (Dec '24) | Top Pick De V3.1 Terminus (Non-reasoning) |
|---|---|---|
| Pricing per 1M tokens | ||
| Input Cost | $0.40/1M | $0.27/1M |
| Output Cost | $0.89/1M | $1.00/1M |
| Blended (3:1) | $0.52/1M | $0.45/1M |
| Specifications | ||
| Organization | DeepSeek | DeepSeek |
| Release Date | Dec 26, 2024 | Sep 22, 2025 |
| Performance & Speed | ||
| Throughput | — | — |
| TTFT | — | — |
| Latency | — | — |
| Composite Indices | ||
| Value Score | 68.2 | 100.0 |
| Reasoning Score | 34.5 | 43.9 |
| Intelligence | 16.5 | 28.5 |
| Coding | 16.4 | 31.9 |
| Math | 26.0 | 53.7 |
| Standard Benchmarks | ||
| GPQA | 55.7% | 75.1% |
| MMLU Pro | 75.2% | 83.6% |
| HLE | 3.6% | 8.4% |
| LiveCodeBench | 35.9% | 52.9% |
| MATH 500 | 88.7% | — |
| AIME 2025 | 26.0% | 53.7% |
| AIME (Original) | 25.3% | — |
| SciCode | 35.4% | 32.1% |
| LCR | 29.0% | 43.3% |
| IFBench | 34.8% | 41.2% |
| TAU-bench v2 | 22.8% | 37.1% |
| TerminalBench Hard | 6.8% | 31.8% |
Key Takeaways
V3.1 Terminus (Non-reasoning) offers the best value at $0.27/1M, making it ideal for high-volume applications and cost-conscious projects.
V3.1 Terminus (Non-reasoning) has the strongest reasoning profile with a 43.9 reasoning score, combining the available reasoning-heavy benchmarks.
V3.1 Terminus (Non-reasoning) reaches a 31.9 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
V3 (Dec '24)
- General-purpose AI
- Versatile applications
V3.1 Terminus (Non-reasoning)
- Cost-sensitive applications
- High-volume processing
- Complex reasoning tasks
- Research & analysis
- Code generation
- Software development
