Dagens Vibes — 29. august 2026

Chris@ChrisGPT · 28. august · 10:26

A new open weight Chinese model has hit the timeline!! This time it’s from Tencent’s Hunyuan team with Hy4 preview. It’s a massive MoE with 770B total parameters / 49B active per token, a 1M context window, and the weights are released under Apache 2.0. Important caveat - Tencent says this is still an early Hy4 checkpoint with more pretraining + post training to come. I actually like how honest their benchmark sheet is. They show plenty of places where the model is still behind despite its size. Hy4 gets 85.4 on Terminal Bench 2.1, basically right in the frontier cluster, and jumps from Hy3’s 28.0 -> 64.3 on DeepSWE. But it’s still behind Kimi K3 at 74.0 and Claude Opus 5 at 74.7 there. On ProgramBench it gets 17.5 vs Claude’s 39.5, SWE Atlas Refactoring 53.3 vs 60.0, and Humanity’s Last Exam 43.4 vs 53.2. It’s a 49B active open weight model that is competitive for its size on a bunch of hard coding/agent benchmark. One thing Tencent also reported on GitHub is they ran a 163 person internal blind eval across 203 engineering tasks where Hy4 slightly beat GLM 5.3 and Kimi K3, which is pretty interesting.

Chubby♨️@kimmonismus · 28. august · 10:15

Most AI benchmarks test whether a model can give the right answer. CommerceAgentBench asks whether an agent can actually finish the work. Accio has open-sourced 107 e-commerce tasks across procurement, product listings, operations, fulfillment and after-sales. Agents work across browsers, email, calendars, documents, APIs and files. Crucially, they are not graded on what they claim to have done. The benchmark verifies what they actually changed, saved or submitted. Take the Gmail procurement case. The agent must search roughly 300 messy emails, identify the real suppliers, reconstruct the latest quotes, compare six Incoterms and four currencies, calculate landed costs and detect payment fraud. Then it must choose a supplier, label the relevant emails, save a reply draft and create a kickoff event. This is the kind of benchmark I find genuinely useful. It measures agents more like workers than chatbots. And the results show why human oversight still matters: the best observed run completed only 66 of 107 tasks, a 61.7% pass rate. And since 2026 is literally the year of agents, this is more important than ever. Accio says the tasks draw on 10M SMB users, 1.6M conversations, 200K agent trajectories and Alibaba’s 27 years of e-commerce experience. The project and task specifications are open source: https://t.co/Tq4UKfJDBb

Ray Fernando@RayFernando1337 · 28. august · 06:01

Factory has a super nova on their hands rn. OMG!! I think these guys are going to crack long running agents by the end of the year. Extreme alpha in this article. “The single agent didn't lack skill. It lacked a standard of completion. An independent standard, authored by the same model, drove the implementation much closer to behavioral parity with the reference. What generalizes to real software work is the need for an external, executable standard of completion - one derived from the outcome, before implementation narrows attention, and kept current until the work meets it.”