LOADING DATASHEET
GROUNDED IN REAL HUMAN CONVERSATIONSYOUR BENCH
Tencent Hy3
◆ OPEN
by Tencent
OPEN WEIGHTSLOCALAPIREASONING
RANKED #96 OF 232 ASSISTANTS · OVERALL #165 OF 6,578 · VIBE SCORE 6.0 · 67 VOICES
Tencent Hy reasoning model for coding, instruction following, and agent tasks
HAPPY10.0
14 mentions
SAD-
0 mentions
VIBE TRAJECTORY
26 weeks · 122 voices
AGING WELL
The crowd is warmer now than it was at the start.
RUN IT ON YOUR OWN MACHINE
Open weights: download Tencent Hy3 once and it's yours. No subscription, no rate limits, works offline.
295B total MoE parameters (21B active), a Q4 quantization is roughly 165 GB to 180 GB
- BARE MINIMUM
Apple Mac Studio (M2 Ultra, 192 GB Unified Memory)
192 GB UNIFIED MEMORY
~$6500 USEDRuns tightly quantized Q4 MoE models in unified memory at low token speed
- SWEET SPOT
Custom Multi-GPU Workstation with 8x NVIDIA RTX 3090 24GB
192 GB VRAM
~$7200 USEDComfortable daily driver running Q4 to Q5 quantization with fast MoE throughput
- FULL POWER
Server node with 8x NVIDIA RTX 4090 24GB
192 GB VRAM
~$16000 NEWFast full-speed Q5 to Q8 inference across tensor-parallel pipelines
LEARN HOW TO SET IT UP
WILL MY BOX RUN TENCENT HY3?
The three cards above are researched picks. Search the machine you actually have — we only say yes if it should reply at a usable speed, not just load the weights and crawl.
HARDWARE PICKS & PRICES ARE RESEARCHED FROM THE LIVE WEB AND REFRESHED AUTOMATICALLY. TREAT THEM AS BALLPARK, NOT GOSPEL.
WHAT CAN YOU ACTUALLY DO WITH IT?
Everyday jobs on each of those rigs: how long Tencent Hy3 takes, and how good the crowd says it is at that kind of work.
| THE JOB | BARE MINIMUMApple Mac Studio (M2 Ultra, 192 GB Unified Memory) | SWEET SPOTCustom Multi-GPU Workstation with 8x NVIDIA RTX 3090 24GB | FULL POWERServer node with 8x NVIDIA RTX 4090 24GB |
|---|---|---|---|
| Write an emaila paragraph or two, drafted from a one-line brief | ~83 S | ~21 S | ~11 S |
| Summarize a documenta long report boiled down to the points that matter | ~4.2 MIN | ~63 S | ~33 S |
| Build a websitea small landing page, markup and styles together | ~22 MIN★★★★★ | ~5.6 MIN★★★★★ | ~3 MIN★★★★★ |
| Build a backendan API with routes, storage and tests | ~69 MIN★★★★★ | ~17 MIN★★★★★ | ~9.3 MIN★★★★★ |
| Build a gamea playable browser game in one file | ~42 MIN★★★★★ | ~10 MIN★★★★★ | ~5.6 MIN★★★★★ |
STARS ARE THE COMMUNITY SCORE FOR THAT KIND OF WORK, DOCKED FOR HOW SQUASHED THE WEIGHTS GET AT EACH TIER. TIMES ARE COMPUTED FROM RESEARCHED SPEEDS. BALLPARK, NOT GOSPEL.
WHAT THE INTERNET SAYS
Not enough discussion about Tencent Hy3 yet to call it. We found 67 posts but people did not say much either way.
64 POSTS · 3 COMMENTS · REDDIT · GITHUB · X · BLUESKY
PEOPLE LOVE
- No fan mail yet. The crowd is still warming up.
PEOPLE COMPLAIN ABOUT
- No complaints yet. The internet is still looking for one. Suspicious.
LOOKING FOR AN ALTERNATIVE?
Other models the crowd has fully reviewed, starting with text models like this one.
GLM 5.3
DeepSeek V4 Flash> SOURCES (12)
- REDDIT · 2026-07-06 · 412 UPVOTESNew open model from Tencent Hy: Hy3 (295B total 21B active - apache 2.0)
New open model from Tencent Hy: Hy3 (295B total 21B active - apache 2.0)
- REDDIT · 2026-04-30 · 318 UPVOTESLocal AI News You Missed - April 2026
Local AI News You Missed - April 2026
- REDDIT · 2026-07-10 · 271 UPVOTESTencent-HY3 is the real deal on 128GB!
Tencent-HY3 is the real deal on 128GB!
- REDDIT · 2026-04-27 · 159 UPVOTESGPT-5.5 improves over GPT-5.4 and overtakes Opus 4.6 to take the 2nd place behind Gemini 3.1 Pro on the Extended NYT Connections Benchmark
GPT-5.5 improves over GPT-5.4 and overtakes Opus 4.6 to take the 2nd place behind Gemini 3.1 Pro on the Extended NYT Connections Benchmark
- REDDIT · 2026-05-05 · 61 UPVOTESUpdate to the LLM Debate Benchmark: GPT-5.5, Grok 4.3, DeepSeek V4 Pro, GLM-5.1, Kimi K2.6, Qwen 3.6 Max Preview, Xiaomi MiMo V2.5 Pro,…
Update to the LLM Debate Benchmark: GPT-5.5, Grok 4.3, DeepSeek V4 Pro, GLM-5.1, Kimi K2.6, Qwen 3.6 Max Preview, Xiaomi MiMo V2.5 Pro, Tencent Hy3 Preview, and Mistral Medium 3.5 High Reasoning added
- REDDIT · 2026-07-11 · 49 UPVOTESHy3 (295B MoE) and NVIDIA Nemotron-Labs-Audex-30B-A3B (audio-capable 30B MoE) GGUF quants
Hy3 (295B MoE) and NVIDIA Nemotron-Labs-Audex-30B-A3B (audio-capable 30B MoE) GGUF quants
- REDDIT · 2026-07-19 · 36 UPVOTES192GB gang - what are you running?
192GB gang - what are you running?
- X · 2026-08-21 · 29 UPVOTESI mapped the labs' free tiers in the article. These platforms made free access a daily habit, not a one-time trial: > To
I mapped the labs' free tiers in the article. These platforms made free access a daily habit, not a one-time trial: > To
- X · 2026-08-24 · 24 UPVOTES🔥 2T Tokens in Just 7 Days! Free Model Campaign Is Still Going Strong! The momentum is buildi
🔥 2T Tokens in Just 7 Days! Free Model Campaign Is Still Going Strong! The momentum is buildi
- X · 2026-08-21 · 20 UPVOTESSmall but important correction: these are NOT "open source AI". Genuinely permissive on the released weights and code: D
Small but important correction: these are NOT "open source AI". Genuinely permissive on the released weights and code: D
- X · 2026-08-22 · 18 UPVOTES🔥 1T Tokens in Just 5 Days! The momentum is unstoppable! total Token throughput has officiall
🔥 1T Tokens in Just 5 Days! The momentum is unstoppable! total Token throughput has officiall
- X · 2026-08-23 · 13 UPVOTES🚀 1T TOKENS IN JUST 5 DAYS! THE MOMENTUM IS UNSTOPPABLE! 🎉 total token throughput has offici
🚀 1T TOKENS IN JUST 5 DAYS! THE MOMENTUM IS UNSTOPPABLE! 🎉 total token throughput has offici