Head-to-head comparison across 3benchmark categories. Overall scores shown here use BenchLM's provisional ranking lane.
Hy3 Preview
62
Ling 2.6 Flash
36
Pick Hy3 Preview if you want the stronger benchmark profile. Ling 2.6 Flash only becomes the better choice if knowledge is the priority or you need the larger 262K context window.
Coding
+33.0 difference
Knowledge
+12.3 difference
Inst. Following
+6.1 difference
Hy3 Preview
Ling 2.6 Flash
$0 / $0
$null / $null
N/A
209.5 t/s
N/A
1.07s
256K
262K
Pick Hy3 Preview if you want the stronger benchmark profile. Ling 2.6 Flash only becomes the better choice if knowledge is the priority or you need the larger 262K context window.
Hy3 Preview is clearly ahead on the provisional aggregate, 62 to 36. The gap is large enough that you do not need to squint at the spreadsheet to see the difference.
Hy3 Preview's sharpest advantage is in coding, where it averages 60 against 27. The single biggest benchmark swing on the page is GPQA, 87.2% to 59%. Ling 2.6 Flash does hit back in knowledge, so the answer changes if that is the part of the workload you care about most.
Hy3 Preview is the reasoning model in the pair, while Ling 2.6 Flash is not. That usually helps on harder chain-of-thought-heavy tests, but it can also mean more latency and more token spend in real use. Ling 2.6 Flash gives you the larger context window at 262K, compared with 256K for Hy3 Preview.
Hy3 Preview is ahead on BenchLM's provisional leaderboard, 62 to 36. The biggest single separator in this matchup is GPQA, where the scores are 87.2% and 59%.
Ling 2.6 Flash has the edge for knowledge tasks in this comparison, averaging 59 versus 46.7. Inside this category, GPQA is the benchmark that creates the most daylight between them.
Hy3 Preview has the edge for coding in this comparison, averaging 60 versus 27. Inside this category, SciCode is the benchmark that creates the most daylight between them.
Hy3 Preview has the edge for instruction following in this comparison, averaging 63.1 versus 57. Inside this category, IFBench 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.