Compare / qwen3-4b-instruct-2507 vs GPT-5.2
qwen3-4b-instruct-2507 vs GPT-5.2
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 | qwen3-4b-instruct-2507 | GPT-5.2 |
|---|---|---|---|
| Artificial Analysisreported | Artificial Analysis Intelligence Indexindex score | 6.70 | 30.4 |
| Artificial Analysisreported | GPQA Diamond% of questions | 51.7% | 90.3% |
| Artificial Analysisreported | Humanity's Last Exam% of questions | 4.5% | 37.7% |
| Artificial Analysisreported | IFBench% of instructions | 33.5% | 75.4% |
| Artificial Analysisreported | Long-context reasoning (AA-LCR)% of questions | 11.3% | 82.7% |
| Artificial Analysisreported | Terminal-Bench Hard% of tasks | 4.5% | 47.0% |
| Artificial Analysisreported | Τ²-bench telecom% of tasks | 26.6% | 84.8% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Irrelevance detection% correct | 84.9% | 87.3% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Memory% correct | 23.9% | 45.8% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Multi-turn tasks% correct | 22.1% | 43.8% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Overall accuracy% correct | 35.7% | 55.9% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Relevance detection% correct | 87.5% | 75.0% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Single-turn calls (curated)% correct | 87.9% | 81.8% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Single-turn calls (user-contributed)% correct | 76.4% | 70.4% |
| Berkeley Function Calling Leaderboard (BFCL) V4reported | Web search% correct | 4.5% | 75.5% |
| UGI Leaderboardreported | Requested-length error% off the requested word count | 8.0% | 76.0% |
| UGI Leaderboardreported | Style adherencescore from 0 to 1 | 0.34 | 0.40 |
| UGI Leaderboardreported | Writing scorescore out of 100 | 29.9 | 49.4 |
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.