what are some good evals for base models? will run! (FYI i expect it to not perform _that_ well since the model card emphasizes they really only pretrained gpt-oss to be good at math/code/reasoning. also my basemodelrecovery process is likely somewhat lossy.)
RESEARCH
-
Running AI Model Evaluations and Performance Benchmarks
By
–
can you give some examples? i bet it doesn't perform better– but i can run the evals!
-
AI Builders Intent: Examining Bias in Tech Criticism Narratives
By
–
My point is that Karen Hao is feeding your ressentiments. Your belief that AI builders care less about the future of humanity than eg. anti tech journalists may be unfounded and clouded by totalizing ideological narratives, which is why you believe the conclusion regardless of
-

Antinatalism prevalence among journalists versus AI researchers
By
–
Given that this headline got cited thousands of times by various outlets, would it make sense for you to set the record straight? And is the incidence of antinatalism really lower among journalists than among AI researchers?
-

GLM-4.5 Analysis: Chinese Open-Source Model Capabilities
By
–
Is the new GLM-4.5 overhyped?
— Louis-François Bouchard 🎥🤖 (@Whats_AI) 13 août 2025
Let’s break it down 👇
What we know:
• From a Chinese lab
• MoE design: GLM-4.5 (≈355B total, 32 experts) & GLM-4.5-Air (≈106B, 12 experts) — similar to Kimi K2
• Pushing open-source AI forward
• SOTA open-model results; reportedly… pic.twitter.com/4RnbS5OFHCIs the new GLM-4.5 overhyped? Let’s break it down What we know: • From a Chinese lab • MoE design: GLM-4.5 (≈355B total, 32 experts) & GLM-4.5-Air (≈106B, 12 experts) — similar to Kimi K2 • Pushing open-source AI forward • SOTA open-model results; reportedly
-

Claude Sonnet 4: 1 Million Context Window, A Leap Forward?
By
–
Breaking news: Claude Sonnet 4 now offers 1 million context! Remains to be seen in actual use, but if the retrieval is as good as Google's, this is truly a leap forward for Claude.
-

India AI Impact Summit 2026 Pre-Summit Events Announced
By
–
Inviting organisations worldwide to host Pre-Summit Events leading up to the India – AI Impact Summit 2026. Join us in advancing conversations on responsible AI, guided by the summit’s three core sutras — People, Planet, Progress — and aligned with the seven key thematic
-

Tencent Launches Hunyuan-Large-Vision Multimodal Model
By
–
Tencent announced Hunyuan-Large-Vision, a multimodal understanding model It excels in complex tasks like visual reasoning, video understanding, and 3D spatial comprehension Also scored 1256 on the LMArena Vision leaderboard, on par with frontier models
-

NewTuring nurtures AI talents benchmark achievement Asia-Pacific
By
–
Congrats @longphan3110 on yet another cool benchmark (after HLE). Long grew up from @newturing (formerly VietAI) working with me & @thtrieu_ on machine translation. Now return back to https://t.co/fRvhGyWatj to help nurture young talents from Asia-Pacific (<12hrs left to apply!) https://t.co/oMgK95hRXG
— Thang Luong (@lmthang) 13 août 2025Congrats @longphan3110 on yet another cool benchmark (after HLE). Long grew up from @newturing (formerly VietAI) working with me & @thtrieu_ on machine translation. Now return back to https://
gstar.newturing.ai to help nurture young talents from Asia-Pacific (<12hrs left to apply!) -
AI addresses India’s stray dog crisis with control and compassion
By
–
Can AI solve India’s stray dog dilemma? In my latest piece, I explore how technology can walk the tightrope between control and compassion.