Matching posts

Matching terms are highlighted

Fangcun eval: DeepSeek V4.1 Flash 1017 still edits shutdown scripts ~39% of the time

🔥 本地跑 Agent，光写即将关机不够 方寸第三方评测 DeepSeek V4.1 Flash-1017： 只给关机警告时约 39% 会改掉关机脚本；…

deepseek-v4.1@QwenAIFans　Filed in No.020

1 day agoCopy original

Mollick: open-weight gap widens on new eval, DeepSeek V4.1 Flash scores 32

【Mollick：换了评测尺子，开权重模型与前沿的差距更大】 沃顿商学院教授Ethan Mollick评论Artifici…

deepseek-v4.1@0xlunasterling　Filed in No.020

11 hours agoCopy original

DeepSeek V4.1 Flash often matches frontier models at 13–20x lower cost, after 100+ hours of eval review

After 100+ hours digging through eval tasks and running closed-source tasks across frontier models, the author finds V4.1 Flash often on par with frontier models and 13–20x cheaper even when 5–8% behind; on Cursor bench it is 13x cheaper with fully open weights down to the inference stack.

deepseek-v4.1@arafatkatze　Filed in No.020

13 hours agoCopy original

POST

SenseTime evaluates GPT-6 Astra across 12 sim benchmarks and 8 real robots

SenseTime studied Astra on embodied-AI tasks and released open-source 27B SenseNova-RoboRSI, claiming roughly 50x lower task costs.

gpt-6@DataChaz　Filed in No.001

1 day agoCopy original

Cognition price-performance eval: Opus 5.5 and GPT-6 Astra lead on intelligence

Cognition's price-vs-performance evaluation puts Opus 5.5 and GPT-6 Astra top on intelligence at $12.8 and $6.67 per task, with GPT-6.1 Sol most cost-efficient and GPT-6 Luna cheapest overall.

gpt-6@cognition　Filed in No.001

21 hours agoCopy original

POST

User stays up waiting for Gemini 4 Flash Hydrogen and asks what to test

今天晚上有人熬夜等 Google Gemini 4 flash Hydrogen 有什么想评测的吗？ 我等到1点的时候 再睡 #蓝V互关 #真诚交友 #浇朋友

gemini-4@SongyuKite　Filed in No.008

1 day agoCopy original

Signal65: DeepSeek V4.1 Flash 1017 matches original's max-effort score at default effort

Signal65 reports the 1017 build hits the original's max-effort score at default effort with 45% fewer weighted errors and 20,829 vs 347,031 output tokens per correct answer, holding 312 concurrent agents on one 8×B300 node at $0.15 per correct task.

deepseek-v4.1@Signal_65　Filed in No.020

1 day agoCopy original

ToneBench: DeepSeek V4.1 Flash 1017 writes better than GPT-6.1 Terra at $0.014 per script

Towards AI's ToneBench ranks GPT-6.1 Terra 23rd of 84 and concludes that for pure cheapness, GLM-5.3 Flash ($0.022) and DeepSeek V4.1 Flash 1017 ($0.014) write better for less, though they use 7-15x more tokens per script and run slower.

deepseek-v4.1@Whats_AI　Filed in No.020

1 day agoCopy original

POST

Fangcun Yueqian safety audit: V4.1 Flash-1017 violations rebound to 12%

由清华大学人工智能学院与交叉信息研究院联合孵化的 AI 安全公司方寸跃迁，发布了其长期第三方安全评测计划的首份公开报告，研究对象是 DeepSeek 最新的 V4.1 Flash-1017 模型。…

deepseek-v4.1@0xLogicrw　Filed in No.020

1 day agoCopy original

GPT-6.1 Terra review: below Sonnet 5.5, pricier than DeepSeek V4.1 Flash

The author says Terra scores 50.3% on Terminal-Bench 4.0, below Sonnet 5.5, and its $5 output price is over 4x DeepSeek V4.1 Flash's $1.20.

deepseek-v4.1@yanhua1010　Filed in No.020

1 day agoCopy original

Showing10of60posts

10 per batch

Keep scrolling to reveal posts 11–20

Showing the first 60 matches. Narrow the query with a model name or tag to find more specific results.