Hy3 preview
Hy3 previewRankingin non-thinking mode?
Alternatives?
CheaperClaude Haiku 5.5max×2.5-30%$0.21Remove sampling parameters+5
BetterGPT-6 Lunalow+26%-30%$0.005Function calling only with none · Chat Completions+3
OpenStep 5 Previewthinking×2.6+400%$1.03Memory (4-bit): 179 GB → 363 GB+3
CheaperMiniMax M3.1 Flashthinking×2.5-8%$0.22Complete model response must be appended
CheaperGLM 5.3 FlashmaxOpen×2.5-17%$0.28Memory (4-bit): 179 GB → 193 GB+5Claude Haiku 5.5max×2.5-30%$0.21Remove sampling parameters+5MiniMax M3.1 Flashthinking×2.5-8%$0.22Complete model response must be appendedGLM 5.3 FlashmaxOpen×2.5-17%$0.28Memory (4-bit): 179 GB → 193 GB+5
Qwen3.8 FlashthinkingOpen×2.3-19%$0.37max_tokens range · QwenCloud+5GPT-6 Lunamax×2.2-30%$0.068Function calling only with none · Chat Completions+3GPT-6 Lunalow+26%-30%$0.005Function calling only with none · Chat Completions+3GPT-5.6 Lunalow+23%+58%$0.01Claude Haiku 5.5low+73%-30%$0.024Remove sampling parameters+5
Ling 3.0 FlashthinkingOpen+18%-68%$0.028
MiMo v2.5thinkingOpen+28%-39%$0.042Memory (4-bit): 179 GB → 186 GBStep 5 PreviewthinkingOpen×2.6+400%$1.03Memory (4-bit): 179 GB → 363 GB+3GLM 5.3 FlashmaxOpen×2.5-17%$0.28Memory (4-bit): 179 GB → 193 GB+5Ling 3.1 FlashOpen×2.4+58%$1.01Memory (4-bit): 179 GB → 343 GBQwen3.8 FlashthinkingOpen×2.3-19%$0.37max_tokens range · QwenCloud+5MiMo v2.6 FlashthinkingOpen×2.2-39%$0.062Memory (4-bit): 179 GB → 186 GB+4Hy3thinkingOpen+48%—$0.072Hy4 previewOpenn/a—$0.83 / $2.50Memory (4-bit): 179 GB → 468 GBHy MT2 1.8BOpenn/a—$0.044 / $0.18Context: 256K → 8KHy MT2 30B A3BOpenn/a—$0.074 / $0.29Context: 256K → 8KHy MT2 7BOpenn/a—$0.074 / $0.29Context: 256K → 8KHunyuan A13B InstructOpenn/a—$0.14 / $0.57Context: 256K → 131KHunyuan A13bn/a——Context: not verifiedAll Tencent models →Run it yourself
estimateBenchmarks
all categoriesFAQ
How much does Hy3 preview cost?›
$0.18 per 1M input tokens and $0.60 per 1M output tokens (cheapest host: GMICloud).
Is Hy3 preview free to use?›
There is no free API; the cheapest paid price is $0.18 per 1M input tokens. The weights are open, so you can run it yourself.
How good is Hy3 preview, and what is it best at?›
#224 of 336 in Chat & Knowledge, #81 of 97 in Coding, #95 of 242 in Agents & Tool Use, #104 of 197 in Writing & Language.
Is Hy3 preview better than MiniMax M2.7?›
On AA Intelligence Index 4.3.2, Hy3 preview scores 22.7 vs 22.8 — about the same.
How can I use Hy3 preview?›
Through the API of GMICloud.
What is the context window of Hy3 preview?›
256K tokens, with up to 236K output tokens.
When was Hy3 preview released? Is there a newer version?›
Released 2026-04-23. It has been superseded by Hy3.
Is Hy3 preview open source?›
Yes, the weights are open.
Can I run Hy3 preview locally?›
The weights are open and can be self-hosted; hardware needs are not confirmed yet.
Can Hy3 preview read images?›
No, it accepts text only.
What is Hy3 preview?›
Hy3 preview is an open-weights language model from Tencent, released Apr 2026.
What are the best alternatives to Hy3 preview?›
Claude Haiku 5.5 (cheaper), GPT-6 Luna (better), Step 5 Preview (open weights).