Head-to-head comparison across 3benchmark categories. Overall scores shown here use BenchLM's provisional ranking lane.
GPT-4.1
58
Qwen3.5 397B
64
Verified leaderboard positions: GPT-4.1 unranked · Qwen3.5 397B #15
Pick Qwen3.5 397B if you want the stronger benchmark profile. GPT-4.1 only becomes the better choice if knowledge is the priority or you need the larger 1M context window.
Coding
+5.7 difference
Knowledge
+1.1 difference
Inst. Following
+5.2 difference
GPT-4.1
Qwen3.5 397B
$2 / $8
$0.6 / $3.6
108 t/s
96 t/s
1.02s
2.44s
1M
128K
Pick Qwen3.5 397B if you want the stronger benchmark profile. GPT-4.1 only becomes the better choice if knowledge is the priority or you need the larger 1M context window.
Qwen3.5 397B is clearly ahead on the provisional aggregate, 64 to 58. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
Qwen3.5 397B's sharpest advantage is in coding, where it averages 60.3 against 54.6. The single biggest benchmark swing on the page is GPQA, 66.3% to 88.4%. GPT-4.1 does hit back in knowledge, so the answer changes if that is the part of the workload you care about most.
GPT-4.1 is also the more expensive model on tokens at $2.00 input / $8.00 output per 1M tokens, versus $0.60 input / $3.60 output per 1M tokens for Qwen3.5 397B. That is roughly 2.2x on output cost alone. GPT-4.1 gives you the larger context window at 1M, compared with 128K for Qwen3.5 397B.
Qwen3.5 397B is ahead on BenchLM's provisional leaderboard, 64 to 58. The biggest single separator in this matchup is GPQA, where the scores are 66.3% and 88.4%.
GPT-4.1 has the edge for knowledge tasks in this comparison, averaging 66.3 versus 65.2. Inside this category, GPQA is the benchmark that creates the most daylight between them.
Qwen3.5 397B has the edge for coding in this comparison, averaging 60.3 versus 54.6. Inside this category, SWE-bench Verified is the benchmark that creates the most daylight between them.
Qwen3.5 397B has the edge for instruction following in this comparison, averaging 92.6 versus 87.4. Inside this category, IFEval is the benchmark that creates the most daylight between them.
For engineers, researchers, and the plain curious — a weekly brief on new models, ranking shifts, and pricing changes.
Free. No spam. Unsubscribe anytime.