Qwen3 235B A22B Instruct 2507
by AlibabaRank #150Score 64.7
Qwen3-235B-A22B-Instruct-2507 is a multilingual, instruction-tuned mixture-of-experts language model based on the Qwen3-235B architecture, with 22B active parameters per forward pass. It is optimized for general-purpose text generation, including instruction following,...
Performance Overview
Score
64.7
Rank
#150
24h Change
+3
7d Change
-13
State
stableConfidence
highSignal Scores
| Signal | Normalized | Weight | Contribution | Freshness |
|---|---|---|---|---|
Capabilities capability | 50.0 | 20% | 10.0 | 2026-11-06T14:59:44.261Z |
Benchmarks benchmark | 64.3 | 30% | 19.3 | 2026-11-06T14:59:44.261Z |
Pricing pricing_tier | 99.2 | 15% | 14.9 | 2026-11-06T14:59:44.261Z |
Context Window context_window | 86.0 | 10% | 8.6 | 2026-11-06T14:59:44.261Z |
Recency recency | 46.7 | 15% | 7.0 | 2026-11-06T14:59:44.261Z |
Output Capacity output_capacity | 81.1 | 10% | 8.1 | 2026-11-06T14:59:44.261Z |
Top Drivers
positive
Benchmarks
positive
Pricing
$0.80/M output tokens
$0.80
neutral
Capabilities
Supports tools, JSON mode, streaming
3/7
positive
Context Window
262K token context window
262K
Capabilities
| Capability | Supported |
|---|---|
| Vision | No |
| Reasoning | No |
| JSON Mode | Yes |
| Streaming | Yes |
| Function Calling | Yes |
| Web Search | No |
Pricing
Input / 1M tokens
$0.14
Output / 1M tokens
$0.80
Context Window
262K
Max Output
118K