Compare / Qwen3 30B A3B Instruct 2507 vs qwen3-4b-instruct-2507
Qwen3 30B A3B Instruct 2507 vs qwen3-4b-instruct-2507
18 results from 3 sources that measured both models. Each row is in the source's own units; there is no overall winner.
| Benchmark | Metric | Qwen3 30B A3B Instruct 2507 | qwen3-4b-instruct-2507 |
|---|---|---|---|
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 7.50 | 6.70 |
| Artificial Analysisreported | GPQA Diamond% of questions | 65.9% | 51.7% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 6.9% | 4.5% |
| Artificial Analysisreported | IFBench% of instructions | 33.1% | 33.5% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 26.3% | 11.3% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 6.1% | 4.5% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 10.2% | 26.6% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Irrelevance detection% correct | 79.9% | 84.9% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Memory% correct | 17.6% | 23.9% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Multi-turn tasks% correct | 30.0% | 22.1% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Overall accuracy% correct | 41.4% | 35.7% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Relevance detection% correct | 93.8% | 87.5% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Single-turn calls (curated)% correct | 88.9% | 87.9% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Single-turn calls (user-contributed)% correct | 78.4% | 76.4% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Web search% correct | 22.5% | 4.5% |
| UGI Leaderboardreported | Requested-length error% off the requested word count | 13.0% | 8.0% |
| UGI Leaderboardreported | Style adherencescore from 0 to 1 | 0.31 | 0.34 |
| UGI Leaderboardreported | Writing scorescore out of 100 | 27.9 | 29.9 |
Bold green marks the better value on that metric. Where a source reports ranges that overlap, the difference may not be meaningful; see the benchmark page for ranges.