Composer 2.5 is exceptionally intelligent & up to 10x more efficient than similarly capable models.

Composer 2.5 is exceptionally intelligent & up to 10x more efficient than similarly capable models.

$84 vs $954 across the same 100 tasks, or ~11x cheaper.

20亿参数对比同体量自回归模型、千亿参数LLaDA2.0，连续路线的scaling曲线健康有效。

ELF用Flow Matching完成生成，仅32个采样步生成质量就超过离散模型1024步结果

训练数据约450亿token，仅为主流方法的十分之一。

Building an AI agent to help clients adjust to changing tax regulations used to take weeks and required teams to switch between multiple tools and chat windows

One senior accountant who spent 180 hours on tax prep last year spent only 15 hours on it this year.

their rate of bug-finding has increased by more than a factor of ten

their rate of bug-finding has increased by more than a factor of ten

$0.07 /Mtok in · $0.014 /Mtok cached

Security work that took hours now takes minutes

This optimization reduced 'write amplification'—the ratio of data written to storage versus the original request—by 20%. It also provided insights for new compiler optimization strategies that reduced the storage footprint of software by nearly 9%.

reduced 'write amplification'—the ratio of data written to storage versus the original request—by 20%

finding 10.4% improvement in routing efficiency over the previous heavily optimized solutions — saving over 15,000 kilometers of distance travelled annually.

humans can do this in well under half an hour.

SubQ Sparse Attention is 52× faster than FlashAttention in our architecture-level comparison, while requiring 63% less compute.

With a research result at 12 million tokens, SubQ's architecture reduces attention compute by almost 1,000x compared to other frontier models.

Overall, it usually takes me about two hours to do this task. If only it were as simple as a single copy and paste, life would be so much easier — or so I thought.

By predicting these unified tokens, it effectively leverages diverse human data to achieve state-of-the-art data efficiency and robust out-of-distribution (OOD) generalization.

They also have the benefits of running on hardware that’s sipping power most of the time, rather than slurping it down in massive data centres.

What used to take reps 5-6 hours a week now runs automatically in the background on every deal.

🔹 **DeepSeek-V4-Flash:** 284B total / 13B active params. Your fast, efficient, and economical choice.

个人学习可能取决于他人行为的主张突出了将学习环境视为一个涉及多个互动参与者的系统的重要性

SSPP # 7.2 Power Usage Effectiveness (PUE) (Electronic Maximum annual weighted average PUE of 1.4 by FY15 )

Google’s ultra-efficient data centers, with a PUE of 1.12, are beating the PUE curve by miles.

When the project is complete later this year (all done while the existing data center remained in operation!), the data center's annual PUE will drop from 1.5 to 1.2, saving 20 percent of its annual electrical cost.

The MGHPCC is targeting a PUE of less than 1.3. A recent report cites typical data center PUEs at 1.9. This means that our facility can expect to