Head-to-head comparison across 4benchmark categories. Overall scores shown here use BenchLM's provisional ranking lane.
Hy3 Preview
62
Kimi K2.5
64
Verified leaderboard positions: Hy3 Preview unranked · Kimi K2.5 #11
Pick Kimi K2.5 if you want the stronger benchmark profile. Hy3 Preview only becomes the better choice if you want the cheaper token bill or you want the stronger reasoning-first profile.
Agentic
+0.2 difference
Coding
+4.2 difference
Knowledge
+18.4 difference
Inst. Following
+30.8 difference
Hy3 Preview
Kimi K2.5
$0 / $0
$0.6 / $3
N/A
45 t/s
N/A
2.38s
256K
256K
Pick Kimi K2.5 if you want the stronger benchmark profile. Hy3 Preview only becomes the better choice if you want the cheaper token bill or you want the stronger reasoning-first profile.
Kimi K2.5 has the cleaner provisional overall profile here, landing at 64 versus 62. It is a real lead, but still close enough that category-level strengths matter more than the headline number.
Kimi K2.5's sharpest advantage is in instruction following, where it averages 93.9 against 63.1. The single biggest benchmark swing on the page is SciCode, 41.2% to 48.7%.
Kimi K2.5 is also the more expensive model on tokens at $0.60 input / $3.00 output per 1M tokens, versus $0.00 input / $0.00 output per 1M tokens for Hy3 Preview. That is roughly Infinityx on output cost alone. Hy3 Preview is the reasoning model in the pair, while Kimi K2.5 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.
Kimi K2.5 is ahead on BenchLM's provisional leaderboard, 64 to 62. The biggest single separator in this matchup is SciCode, where the scores are 41.2% and 48.7%.
Kimi K2.5 has the edge for knowledge tasks in this comparison, averaging 65.1 versus 46.7. Inside this category, HLE is the benchmark that creates the most daylight between them.
Kimi K2.5 has the edge for coding in this comparison, averaging 64.2 versus 60. Inside this category, SciCode is the benchmark that creates the most daylight between them.
Kimi K2.5 has the edge for agentic tasks in this comparison, averaging 54.6 versus 54.4. Inside this category, Terminal-Bench 2.0 is the benchmark that creates the most daylight between them.
Kimi K2.5 has the edge for instruction following in this comparison, averaging 93.9 versus 63.1. Hy3 Preview stays close enough that the answer can still flip depending on your workload.
Estimates at 50,000 req/day · 1000 tokens/req average.
For engineers, researchers, and the plain curious — a weekly brief on new models, ranking shifts, and pricing changes.
Free. No spam. Unsubscribe anytime.