Model Comparison
V3.1 Terminus (Reasoning)
vs. Qwen3.5 397B A17B (Non-reasoning)
Comparing 2 AI models · 10 benchmarks · DeepSeek, Alibaba
Recommended Pick
Al Qwen3.5 397B A17B (Non-reasoning) 11 metric wins Strongest on: Value, Input price, Blended price
Best Value
Qwen3.5 397B A17B (Non-reasoning)
100.0 value score
48.3 reasoning / $1.35/1M
Lowest Price
Qwen3.5 397B A17B (Non-reasoning)
$0.60/1M input price
Best Reasoning
V3.1 Terminus (Reasoning)
61.5 reasoning score
Blends available reasoning benchmarks
Best for Coding
Qwen3.5 397B A17B (Non-reasoning)
37.4 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
Qwen3.5 397B A17B (Non-reasoning) has the strongest quality-to-price mix at 100.0 out of 100 value points.
Price gap
Qwen3.5 397B A17B (Non-reasoning) is 2.7x cheaper on input tokens than V3.1 Terminus (Reasoning).
Reasoning gap
V3.1 Terminus (Reasoning) leads Qwen3.5 397B A17B (Non-reasoning) by 13.2 points on reasoning.
Coding gap
Qwen3.5 397B A17B (Non-reasoning) leads V3.1 Terminus (Reasoning) by 3.7 points on coding.
Top-pick rationale
Qwen3.5 397B A17B (Non-reasoning) wins 11 measurable categories, including Value, Input price, Blended price, Throughput.
Response Face-Off
Run one prompt through the selected models and compare response quality with live speed and cost context.
V3.1 Terminus (Reasoning)
DeepSeek
TTFT
—
Time
—
tok/s
—
Tokens
—
Cost
—
Qwen3.5 397B A17B (Non-reasoning)
Alibaba
TTFT
—
Time
—
tok/s
—
Tokens
—
Cost
—
Which answer was more useful?
Full Comparison
| Metric | De V3.1 Terminus (Reasoning) | Top Pick Al Qwen3.5 397B A17B (Non-reasoning) |
|---|---|---|
| Pricing per 1M tokens | ||
| Input Cost | $1.64/1M | $0.60/1M |
| Output Cost | $2.75/1M | $3.60/1M |
| Blended (3:1) | $1.91/1M | $1.35/1M |
| Specifications | ||
| Organization | DeepSeek | Alibaba |
| Release Date | Sep 22, 2025 | Feb 16, 2026 |
| Performance & Speed | ||
| Throughput | — | 52.9 tok/s |
| TTFT | — | 1934ms |
| Latency | — | 1934ms |
| Composite Indices | ||
| Value Score | 89.8 | 100.0 |
| Reasoning Score | 61.5 | 48.3 |
| Intelligence | 33.9 | 40.1 |
| Coding | 33.7 | 37.4 |
| Math | 89.7 | — |
| Standard Benchmarks | ||
| GPQA | 79.2% | 86.1% |
| MMLU Pro | 85.1% | — |
| HLE | 15.2% | 18.8% |
| LiveCodeBench | 79.8% | — |
| AIME 2025 | 89.7% | — |
| SciCode | 40.6% | 41.1% |
| LCR | 65.0% | 58.0% |
| IFBench | 57.0% | 51.6% |
| TAU-bench v2 | 37.1% | 83.9% |
| TerminalBench Hard | 30.3% | 35.6% |
Key Takeaways
Qwen3.5 397B A17B (Non-reasoning) offers the best value at $0.60/1M, making it ideal for high-volume applications and cost-conscious projects.
V3.1 Terminus (Reasoning) has the strongest reasoning profile with a 61.5 reasoning score, combining the available reasoning-heavy benchmarks.
Qwen3.5 397B A17B (Non-reasoning) reaches a 37.4 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
V3.1 Terminus (Reasoning)
- Complex reasoning tasks
- Research & analysis
Qwen3.5 397B A17B (Non-reasoning)
- Cost-sensitive applications
- High-volume processing
- Code generation
- Software development
