Benchmark: Tencent Hy4 Preview Fixes 17 Bugs for Just $3.13
RiqueRostirolla · x · 2026-08-28
A benchmark of Tencent's Hy4 preview model on real-world code bug fixing.
Test Setup:
- 2 real code repositories.
- 105 hard-planted bugs.
- Task: Find and fix as many bugs as possible.
Results (Fixed / Cost):
- Fable 5 (max): 29/105, $104.49
- Opus 5 (max): 27/105, $51.33
- Qwen3.8-Max (xhigh): 18/105, $31.19
- Hy4 preview (default): 17/105, $3.13
- Gemini 3.7 Flash (high): 16/105, $6.36
Takeaway: Hy4 preview achieved performance comparable to top models (Qwen/Gemini) at a fraction of the cost ($3.13), proving to be a cost-effective and solid model.
Related event: Tencent Hunyuan Hy4 Fixes 17 Bugs for $3.13 in Real-Repo Blind Test(4 posts)→
More from coding & agent
- Marc Lou praises PostShiba, a cute indie alternative to AWS SES for email infra — marclou · 2026-10-16
- RExBench: best coding agent completes only 33% of autonomous research extensions — najoungkim · 2026-10-16
- Wiring Pstack skills from a Grok bot into Hermes agent loops — alexcovo_eth · 2026-10-16
- past.dev launches long-term memory API for AI agents, tops BEAM public benchmark — thisiskp_ · 2026-10-16
- Google's CLASHarness benchmarks multi-agent conflicts in computer use, finds frontier VLMs struggle — google · 2026-10-16
- mitsuhiko's AI agent solved multiplayer with PeerJS before he could explain it — mitsuhiko · 2026-10-16