Model Comparison
Qwen3 235B A22B 2507 Instruct
vs. Qwen3.6 35B A3B (Non-reasoning)
Comparing 2 AI models · 12 benchmarks · Alibaba
Recommended Pick
Al Qwen3 235B A22B 2507 Instruct 10 metric wins Strongest on: Value, Input price, Output price
Best Value
Qwen3 235B A22B 2507 Instruct
100.0 value score
60.6 reasoning / $0.36/1M
Lowest Price
Qwen3 235B A22B 2507 Instruct
$0.20/1M input price
Best Reasoning
Qwen3 235B A22B 2507 Instruct
60.6 reasoning score
Blends available reasoning benchmarks
Best for Coding
Qwen3 235B A22B 2507 Instruct
22.1 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 235B A22B 2507 Instruct has the strongest quality-to-price mix at 100.0 out of 100 value points.
Price gap
Qwen3 235B A22B 2507 Instruct is 1.9x cheaper on input tokens than Qwen3.6 35B A3B (Non-reasoning).
Speed gap
Qwen3.6 35B A3B (Non-reasoning) generates about 2.6x as many tokens per second as Qwen3 235B A22B 2507 Instruct.
Reasoning gap
Qwen3 235B A22B 2507 Instruct leads Qwen3.6 35B A3B (Non-reasoning) by 18.7 points on reasoning.
Coding gap
Qwen3 235B A22B 2507 Instruct leads Qwen3.6 35B A3B (Non-reasoning) by 4.5 points on coding.
Response Face-Off
Run one prompt through the selected models and compare response quality with live speed and cost context.
Qwen3 235B A22B 2507 Instruct
Alibaba
TTFT
—
Time
—
tok/s
—
Tokens
—
Cost
—
Qwen3.6 35B A3B (Non-reasoning)
Alibaba
TTFT
—
Time
—
tok/s
—
Tokens
—
Cost
—
Which answer was more useful?
Full Comparison
| Metric | Top Pick Al Qwen3 235B A22B 2507 Instruct | Al Qwen3.6 35B A3B (Non-reasoning) |
|---|---|---|
| Pricing per 1M tokens | ||
| Input Cost | $0.20/1M | $0.38/1M |
| Output Cost | $0.82/1M | $2.25/1M |
| Blended (3:1) | $0.36/1M | $0.84/1M |
| Specifications | ||
| Organization | Alibaba | Alibaba |
| Release Date | Jul 21, 2025 | Apr 16, 2026 |
| Performance & Speed | ||
| Throughput | 54.0 tok/s | 143.1 tok/s |
| TTFT | 1219ms | 1307ms |
| Latency | 1219ms | 1307ms |
| Composite Indices | ||
| Value Score | 100.0 | 29.2 |
| Reasoning Score | 60.6 | 41.9 |
| Intelligence | 25.0 | 31.5 |
| Coding | 22.1 | 17.6 |
| Math | 71.7 | — |
| Standard Benchmarks | ||
| GPQA | 75.3% | 81.7% |
| MMLU Pro | 82.8% | — |
| HLE | 10.6% | 12.5% |
| LiveCodeBench | 52.4% | — |
| MATH 500 | 98.0% | — |
| AIME 2025 | 71.7% | — |
| AIME (Original) | 71.7% | — |
| SciCode | 36.0% | 1.3% |
| LCR | 31.2% | 56.7% |
| IFBench | 46.1% | 36.2% |
| TAU-bench v2 | 33.3% | 85.1% |
| TerminalBench Hard | 15.2% | 25.8% |
Key Takeaways
Qwen3 235B A22B 2507 Instruct offers the best value at $0.20/1M, making it ideal for high-volume applications and cost-conscious projects.
Qwen3 235B A22B 2507 Instruct has the strongest reasoning profile with a 60.6 reasoning score, combining the available reasoning-heavy benchmarks.
Qwen3 235B A22B 2507 Instruct reaches a 22.1 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
Qwen3 235B A22B 2507 Instruct
- Cost-sensitive applications
- High-volume processing
- Complex reasoning tasks
- Research & analysis
- Code generation
- Software development
Qwen3.6 35B A3B (Non-reasoning)
- General-purpose AI
- Versatile applications
