ModelsCompareBest forBenchmarksStatusPricingAPI
Compare / head-to-head

Hy3vsLongCat Flash Chat

Hy3 leads 3 of 3 shared benchmarks. Hy3 has the larger context window (256K tokens).

Benchmarks from cited public sources; pricing from official pages; status from official provider feeds.

Shared benchmarks
3 – 0
Hy3 leads
Cheaper per token
list price, input + output
Larger context
Hy3
256K tokens
Provider uptime (30d)
— · —
Tencent Hunyuan · Meituan LongCat
Hy3
Tencent Hunyuan · released 2026-07-06
text
Context
256K
Max out
Input /1M
Output /1M
Cached /1M
Scores
41 · 8 core
LongCat Flash Chat
Meituan LongCat · released 2025-09-01
text
Context
128K
Max out
Input /1M
Output /1M
Cached /1M
Scores
31 · 9 core
Quality

Benchmark matrix

BenchmarkHy3LongCat Flash ChatΔEdge
Reported by both · 3
SWE-bench Verified78% 60.4% +17.6 ptHy3
GPQA Diamond90.4% 73.23% +17.2 ptHy3
LMArena Elo1440.6 1422.9 +17.7Hy3
Only Hy3 reports · 38
Humanity's Last Exam (no tools)37% not reported
AA-LCR73.4% not reported
Apex-Agent25.6% not reported
ArxivMath52.2% not reported
BrowseComp84.2% not reported
CL-bench23.8% not reported
CL-bench Life17% not reported
ClawEval68.5% not reported
CMT-Benchmark37.8% not reported
DeepSearchQA91% not reported
DeepSWE28% not reported
e-bench50.2% not reported
FrontierScience-Olympiad74.8% not reported
FrontierScience-Research21.3% not reported
HLE (with tools, text-only)53.2% not reported
HorizonMath7.1% not reported
Hy-Backend 2.025% not reported
Hy-CompanyBench41.7% not reported
Hy-Euler Pro24.2% not reported
Hy-FinModelBench69% not reported
Hy-Math60.9% not reported
Hy-SkillsWorld45.8% not reported
Hy-SWE Max49% not reported
IMOAnswerBench90% not reported
MathArena Apex38.7% not reported
MCP Atlas79.1% not reported
NL2Repo45.6% not reported
PHYBench77.4% not reported
ProdBench23% not reported
SkillsBench55.3% not reported
SuperChem54.9% not reported
SWE-bench Multilingual75.8% not reported
SWE-bench Pro57.9% not reported
Terminal-Bench 2.171.7% not reported
Toolathlon48.5% not reported
USAMO 202672% not reported
WideSearch76.4% not reported
WildClawBench53.6% not reported
Only LongCat Flash Chat reports · 28
MMLUnot reported89.71%
MMLU-Pronot reported82.68%
AIME 2025not reported61.25%
AceBenchnot reported76.1%
AIME 2024not reported70.42%
ArenaHard-V2not reported86.5%
BeyondAIMEnot reported43%
C-Evalnot reported90.44%
CMMLUnot reported84.34%
COLLIEnot reported57.1%
DROPnot reported79.06%
GraphWalks-128knot reported51.05%
HumanEval+not reported88.41%
IFEvalnot reported89.65%
LiveCodeBenchnot reported48.02%
MATH500not reported96.4%
MBPP+not reported79.63%
Meeseeks-zhnot reported43.03%
Safety: Criminalnot reported91.24%
Safety: Harmfulnot reported83.98%
Safety: Misinformationnot reported81.72%
Safety: Privacynot reported93.98%
Tau2-Bench (airline)not reported58%
Tau2-Bench (retail)not reported71.27%
Tau2-Bench (telecom)not reported73.68%
TerminalBenchnot reported39.51%
VitaBenchnot reported24.3%
ZebraLogicnot reported89.3%
Scores tagged "epoch" or "matharena" are independent runs, used only where the lab has not published its own; ⚠ marks rows MathArena flags as released after the competition. Higher is better on every row. Δ is Hy3 minus LongCat Flash Chat in the benchmark's own unit. "Not reported" means the lab has not published that figure; it is not a zero. ↗ opens the source.
Specs & pricing

Side by side

SpecHy3LongCat Flash ChatEdge
Context window256K tokens128K tokensHy3
Max output
Input price / 1M
Output price / 1M
Cached input / 1M
Input + output / 1M
Lower is cheaper. List prices; batch, tool and regional fees excluded.
ModalitiestexttextTie
Released2026-07-062025-09-01
Cited benchmark scores4131
Reliability

Provider status

All providers
More matchups

LongCat Flash Chat vs …

Built by Respan
Which one wins on your data?

Public benchmarks are a starting point. Run Hy3 and LongCat Flash Chat on your own prompts with Respan evals, or route to either through one gateway key with automatic failover.