Macaron V1 Venti posts strong agent and coding benchmark scores across the board
teortaxesTex · x · 2026-07-22
Macaron V1 Venti is shown beating or matching several larger models on a benchmark table across chat, agent, coding, and UI tasks, with particularly strong scores on PinchBench, TerminalBench 2.1, and UI4ABench.
The post argues that Chinese models are still undertrained and short on RL, so relatively modest post-training can still push them further. It also says this should lower confidence in the usefulness of large-scale distillation alone.
Notable scores in the image
- ChatBench: 58.3
- LivingBench: 64.0
- PinchBench: 94.0
- TerminalBench 2.1: 87.6
- DeepSWE: 58.4
- UI4ABench: 87.8
More from Models
- GPT-6 Astra beats Factorio with enemies in 44 in-game hours at ~$4,500 API cost — liminal_bardo · 2026-09-11
- giffmana: the env being used in training is part of the point — giffmana · 2026-09-11
- awesome-llm-leaderboards: an open-source directory of LLM leaderboards, pricing tables, comparison tools — Last_Establishment_1 · 2026-09-11
- Anthropic claims it works to keep eval environments unidentifiable to models — MaxKannen · 2026-09-11
- Nex N2.5 Pro released on Hugging Face with 407GB of weights — jinnyjuice · 2026-09-11
- RoMa v2 image matching model unveiled in the usual black poster — ducha_aiki · 2026-09-11